Decoupling or displacement? The GDP elasticity of consumption-based versus territorial CO2 emissions, 1990-2022
Paper
Decoupling or Displacement? The GDP Elasticity of Consumption-Based versus Territorial CO₂ Emissions, 1990–2022
Abstract
If rich countries decoupled through carbon imports, consumption-based CO₂ should be more GDP-elastic than territorial CO₂. On 119 countries, the top income tercile's 1990–2007 consumption–territorial wedge elasticity is 0.027 (0.117); post hoc variants (without oil exporters and transition economies; 1990 net carbon importers) give 0.305 (0.234) and 0.105 (0.120). None is significant; sizeable displacement cannot be excluded. Dropping Malta post hoc leaves 2007–2022 slopes small and insignificant; no group's slope differs significantly between periods. Aggregate wedges rose 0.052–0.067 log points over 1990–2007 and fell 0.018–0.040 between the 2007 and 2017 benchmarks, significantly in weighted regressions only for 1997 net importers.
JEL: Q54, Q56, F18, O44
§Ask about this paper
Grounded in this runQuestions are answered from this run's record — the paper, the executed analysis, the data provenance and the referee report. Nothing else is consulted.
This run is finished, so the recorded paper cannot be rewritten. Questions are still answered from its record.
Pipeline
1.Data profile
Profiled 50,411 rows and 79 columns; description drafted. Confirmed by the owner.
- File
- owid-co2-data.csv
- Size
- 32 MB · csv
- Rows
- 50,411
- Columns
- 79
- Checksum
- sha256 7ec10de1e9502c20c70bfff54476d7893922f7594aa299d9291d87fe49f3e342
- Supplied
- 29 Oct 2026, 14:35
Description
Confirmed by the owner on 29 Oct 2026, 14:36.
- One row is
- One row is one geographic entity in one calendar year (country-year panel, 1750-2024). Entities are 218 countries/territories with an ISO alpha-3 code plus 36 entities without one: 34 aggregates or non-territorial entities, namely aggregates (World, continents with and without the '(GCP)' suffix, EU-27/EU-28, income groups, OECD/Non-OECD, 'Asia (excl. China and India)', 'North America (excl. USA)' etc.), non-territorial sources (International aviation, International shipping, Kuwaiti Oil Fires), plus two places without ISO codes (Kosovo, Ryukyu Islands). (country, year) is unique: no duplicate pairs. The file appears to be Our World in Data's CO2 and greenhouse-gas emissions dataset (owid-co2-data.csv), as supplied by the user and not independently verified; column meanings and units below follow the public OWID codebook (owid-co2-codebook.csv), which matches every column name in the file.
- Subject
- Economics of carbon emissions: national CO2 and greenhouse-gas emissions, GDP, energy use and trade-embodied (consumption-based) emissions
- Keys
- country, year
- Time variables
- year
| Column | Meaning | Unit | Role |
|---|---|---|---|
| country | Name of the geographic entity: a sovereign country or territory, or an aggregate (continent, income group, EU, 'World', GCP regional groupings, International aviation/shipping, Kuwaiti Oil Fires). | none | identifier |
| year | Calendar year of observation, 1750-2024. | year | time |
| iso_code | ISO 3166-1 alpha-3 code. Present for 218 countries/territories. Missing for 36 entities: 34 aggregates or non-territorial entities (World, continents, GCP regions, EU, income groups, OECD, International aviation/shipping, Kuwaiti Oil Fires) plus two places without an ISO code (Kosovo, Ryukyu Islands). Non-missing iso_code is therefore the natural filter for country-level analysis, at the cost of dropping Kosovo. | none | identifier |
| population | Total population of the entity in that year, compiled by OWID from several sources (as given in the file, 1750-2024). | people | measure |
| gdp | Real GDP, adjusted for inflation and differences in living costs (PPP), from the Maddison Project Database 2023. Covers 1820-2022 at most; no values for 2023-2024. | international-$ at 2011 prices | measure |
| cement_co2 | Annual emissions of carbon dioxide (CO₂) from cement, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| cement_co2_per_capita | Annual emissions of carbon dioxide (CO₂) from cement, measured in tonnes per person. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| co2 | Annual territorial (production-based) CO2 emissions from fossil fuels and industry (coal, oil, gas, flaring, cement, other industry), excluding land-use change. | million tonnes CO2 (Mt) | measure |
| co2_growth_abs | Annual growth in total emissions of carbon dioxide (CO₂), excluding land-use change, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| co2_growth_prct | Annual percentage growth in total emissions of carbon dioxide (CO₂), excluding land-use change. Source per OWID codebook: Global Carbon Budget (2025). | % | measure |
| co2_including_luc | Annual total emissions of carbon dioxide (CO₂), including land-use change, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| co2_including_luc_growth_abs | Annual growth in total emissions of carbon dioxide (CO₂), including land-use change, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| co2_including_luc_growth_prct | Annual percentage growth in total emissions of carbon dioxide (CO₂), including land-use change. Source per OWID codebook: Global Carbon Budget (2025). | % | measure |
| co2_including_luc_per_capita | Annual emissions of carbon dioxide (CO₂), including land-use change, measured in tonnes per person. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| co2_including_luc_per_gdp | Annual total emissions of carbon dioxide (CO₂), including land-use change, measured in kilograms per dollar of GDP (2011 international-$). Source per OWID codebook: Global Carbon Budget (2025). | kilograms per international-$ (kg/$) | measure |
| co2_including_luc_per_unit_energy | Carbon dioxide (CO₂) emitted per unit of primary energy consumption, measured in grams of CO₂ per kilowatt-hour. Land-use change emissions are included. Source per OWID codebook: Global Carbon Budget (2025). | grams per kilowatt-hour (g/kWh) | measure |
| co2_per_capita | Carbon dioxide (CO₂) emissions from burning fossil fuels and industrial processes. This includes emissions from transport, electricity generation, and heating, but not land-use change. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| co2_per_gdp | Production-based CO2 (excl. land-use change) per unit of GDP; equals co2*1e9/gdp (verified in the file). | kg CO2 per international-$ (2011 prices) | measure |
| co2_per_unit_energy | Carbon dioxide (CO₂) emitted per unit of primary energy consumption, measured in grams of CO₂ per kilowatt-hour. Source per OWID codebook: Global Carbon Budget (2025). | grams per kilowatt-hour (g/kWh) | measure |
| coal_co2 | Annual emissions of carbon dioxide (CO₂) from coal, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| coal_co2_per_capita | Annual emissions of carbon dioxide (CO₂) from coal, measured in tonnes per person. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| consumption_co2 | Consumption-based (trade-adjusted) CO2 emissions: territorial emissions plus emissions embodied in imports minus those embodied in exports. Available 1990-2023 for about 120 countries. | million tonnes CO2 (Mt) | measure |
| consumption_co2_per_capita | Annual consumption-based emissions of carbon dioxide (CO₂), measured in tonnes per person. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| consumption_co2_per_gdp | Annual consumption-based emissions of carbon dioxide (CO₂), measured in kilograms per dollar of GDP (2011 international-$). Source per OWID codebook: Global Carbon Budget (2025). | kilograms per international-$ (kg/$) | measure |
| cumulative_cement_co2 | Cumulative emissions of carbon dioxide (CO₂) from cement since the first year of available data, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). Running sum since the first year with data for the entity, so it depends on when coverage begins. | million tonnes (Mt) | measure |
| cumulative_co2 | Total cumulative emissions of carbon dioxide (CO₂), excluding land-use change, since the first year of available data, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). Running sum since the first year with data for the entity, so it depends on when coverage begins. | million tonnes (Mt) | measure |
| cumulative_co2_including_luc | Total cumulative emissions of carbon dioxide (CO₂), including land-use change, since the first year of available data, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). Running sum since the first year with data for the entity, so it depends on when coverage begins. | million tonnes (Mt) | measure |
| cumulative_coal_co2 | Cumulative emissions of carbon dioxide (CO₂) from coal since the first year of available data, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). Running sum since the first year with data for the entity, so it depends on when coverage begins. | million tonnes (Mt) | measure |
| cumulative_flaring_co2 | Cumulative emissions of carbon dioxide (CO₂) from flaring since the first year of available data, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). Running sum since the first year with data for the entity, so it depends on when coverage begins. | million tonnes (Mt) | measure |
| cumulative_gas_co2 | Cumulative emissions of carbon dioxide (CO₂) from gas since the first year of available data, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). Running sum since the first year with data for the entity, so it depends on when coverage begins. | million tonnes (Mt) | measure |
| cumulative_luc_co2 | Cumulative emissions of carbon dioxide (CO₂) from land-use change since the first year of available data, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). Running sum since the first year with data for the entity, so it depends on when coverage begins. | million tonnes (Mt) | measure |
| cumulative_oil_co2 | Cumulative emissions of carbon dioxide (CO₂) from oil since the first year of available data, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). Running sum since the first year with data for the entity, so it depends on when coverage begins. | million tonnes (Mt) | measure |
| cumulative_other_co2 | Cumulative emissions of carbon dioxide (CO₂) from other industry sources since the first year of available data, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). Running sum since the first year with data for the entity, so it depends on when coverage begins. | million tonnes (Mt) | measure |
| energy_per_capita | Primary energy consumption per person. Available from 1965. Source per OWID codebook: U.S. EIA and Energy Institute Statistical Review. | kWh per person | measure |
| energy_per_gdp | Primary energy consumption per unit of GDP (energy intensity of output). Available from 1965 to 2022 (GDP ends in 2022). Source per OWID codebook: U.S. EIA, Energy Institute, Maddison Project. | kWh per international-$ (2011 prices) | measure |
| flaring_co2 | Annual emissions of carbon dioxide (CO₂) from flaring, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| flaring_co2_per_capita | Annual emissions of carbon dioxide (CO₂) from flaring, measured in tonnes per person. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| gas_co2 | Annual emissions of carbon dioxide (CO₂) from gas, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| gas_co2_per_capita | Annual emissions of carbon dioxide (CO₂) from gas, measured in tonnes per person. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| ghg_excluding_lucf_per_capita | Annual total greenhouse-gas emissions excluding LUCF per person, in CO2-equivalents (GWP100) (total_ghg_excluding_lucf / population). Source per OWID codebook: Jones et al., National contributions to climate change (2025). | t CO2-equivalent per person | measure |
| ghg_per_capita | Annual total greenhouse-gas emissions including LUCF per person, in CO2-equivalents (GWP100) (total_ghg / population). Source per OWID codebook: Jones et al., National contributions to climate change (2025). | t CO2-equivalent per person | measure |
| land_use_change_co2 | Annual emissions of carbon dioxide (CO₂) from land-use change, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| land_use_change_co2_per_capita | Annual emissions of carbon dioxide (CO₂) from land-use change, measured in tonnes per person. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| methane | Annual methane (CH4) emissions, including land use, converted to CO2-equivalents with 100-year global warming potentials (GWP100). Source per OWID codebook: Jones et al., National contributions to climate change (2025). | million tonnes CO2-equivalent (GWP100) | measure |
| methane_per_capita | Annual methane (CH4) emissions per person, in CO2-equivalents (GWP100). Source per OWID codebook: Jones et al., National contributions to climate change (2025). | t CO2-equivalent per person | measure |
| nitrous_oxide | Annual nitrous oxide (N2O) emissions, including land use, converted to CO2-equivalents (GWP100). Source per OWID codebook: Jones et al., National contributions to climate change (2025). | million tonnes CO2-equivalent (GWP100) | measure |
| nitrous_oxide_per_capita | Annual nitrous oxide (N2O) emissions per person, in CO2-equivalents (GWP100). Source per OWID codebook: Jones et al., National contributions to climate change (2025). | t CO2-equivalent per person | measure |
| oil_co2 | Annual emissions of carbon dioxide (CO₂) from oil, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| oil_co2_per_capita | Annual emissions of carbon dioxide (CO₂) from oil, measured in tonnes per person. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| other_co2_per_capita | Annual emissions of carbon dioxide (CO₂) from other industry sources, measured in tonnes per person. Source per OWID codebook: Global Carbon Budget (2025). | tonnes per person (t/person) | measure |
| other_industry_co2 | Annual emissions of carbon dioxide (CO₂) from other industry sources, measured in million tonnes. Source per OWID codebook: Global Carbon Budget (2025). | million tonnes (Mt) | measure |
| primary_energy_consumption | Annual primary energy consumption, all sources. Available from 1965. Source per OWID codebook: U.S. EIA International Energy Data and Energy Institute Statistical Review of World Energy. | terawatt-hours (TWh) | measure |
| share_global_cement_co2 | Annual emissions of carbon dioxide (CO₂) from cement, measured as a percentage of global emissions of CO₂ from cement in the same year. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. | % | measure |
| share_global_co2 | Annual total emissions of carbon dioxide (CO₂), excluding land-use change, measured as a percentage of global emissions of CO₂ in the same year. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. | % | measure |
| share_global_co2_including_luc | Annual total emissions of carbon dioxide (CO₂), including land-use change, measured as a percentage of global total emissions of CO₂ in the same year. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. | % | measure |
| share_global_coal_co2 | Annual emissions of carbon dioxide (CO₂) from coal, measured as a percentage of global emissions of CO₂ from coal in the same year. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. | % | measure |
| share_global_cumulative_cement_co2 | Cumulative emissions of carbon dioxide (CO₂) from cement since the first year of available data, measured as a percentage of global cumulative emissions of CO₂ from cement. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. Running sum since the first year with data for the entity, so it depends on when coverage begins. | % | measure |
| share_global_cumulative_co2 | Total cumulative emissions of carbon dioxide (CO₂), excluding land-use change, since the first year of available data, measured as a percentage of global total cumulative emissions of CO₂. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. Running sum since the first year with data for the entity, so it depends on when coverage begins. | % | measure |
| share_global_cumulative_co2_including_luc | Total cumulative emissions of carbon dioxide (CO₂), including land-use change, since the first year of available data, measured as a percentage of global total cumulative emissions of CO₂ (including land-use change). Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. Running sum since the first year with data for the entity, so it depends on when coverage begins. | % | measure |
| share_global_cumulative_coal_co2 | Cumulative emissions of carbon dioxide (CO₂) from coal since the first year of available data, measured as a percentage of global cumulative emissions of CO₂ from coal. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. Running sum since the first year with data for the entity, so it depends on when coverage begins. | % | measure |
| share_global_cumulative_flaring_co2 | Cumulative emissions of carbon dioxide (CO₂) from flaring since the first year of available data, measured as a percentage of global cumulative emissions of CO₂ from flaring. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. Running sum since the first year with data for the entity, so it depends on when coverage begins. | % | measure |
| share_global_cumulative_gas_co2 | Cumulative emissions of carbon dioxide (CO₂) from gas since the first year of available data, measured as a percentage of global cumulative emissions of CO₂ from gas. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. Running sum since the first year with data for the entity, so it depends on when coverage begins. | % | measure |
| share_global_cumulative_luc_co2 | Cumulative emissions of carbon dioxide (CO₂) from land-use change since the first year of available data, measured as a percentage of global cumulative emissions of CO₂ from land-use change. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. Running sum since the first year with data for the entity, so it depends on when coverage begins. | % | measure |
| share_global_cumulative_oil_co2 | Cumulative emissions of carbon dioxide (CO₂) from oil since the first year of available data, measured as a percentage of global cumulative emissions of CO₂ from oil. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. Running sum since the first year with data for the entity, so it depends on when coverage begins. | % | measure |
| share_global_cumulative_other_co2 | Cumulative emissions of carbon dioxide (CO₂) from other industry sources since the first year of available data, measured as a percentage of global cumulative emissions of CO₂ from other industry sources. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. Running sum since the first year with data for the entity, so it depends on when coverage begins. | % | measure |
| share_global_flaring_co2 | Annual emissions of carbon dioxide (CO₂) from flaring, measured as a percentage of global emissions of CO₂ from flaring in the same year. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. | % | measure |
| share_global_gas_co2 | Annual emissions of carbon dioxide (CO₂) from gas, measured as a percentage of global emissions of CO₂ from gas in the same year. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. | % | measure |
| share_global_luc_co2 | Annual emissions of carbon dioxide (CO₂) from land-use change, measured as a percentage of global emissions of CO₂ from land-use change in the same year. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. | % | measure |
| share_global_oil_co2 | Annual emissions of carbon dioxide (CO₂) from oil, measured as a percentage of global emissions of CO₂ from oil in the same year. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. | % | measure |
| share_global_other_co2 | Annual emissions of carbon dioxide (CO₂) from other industry sources, measured as a percentage of global emissions of CO₂ from other industry sources in the same year. Source per OWID codebook: Global Carbon Budget (2025). Share of the World row in the same year; values ~100 occur on aggregate rows. | % | measure |
| share_of_temperature_change_from_ghg | The entity's contribution to global warming from all three gases as a percentage of the World's total warming in that year. Source per OWID codebook: Jones et al., National contributions to climate change (2025). | % | measure |
| temperature_change_from_ch4 | The entity's contribution to the change in global mean surface temperature since pre-industrial times caused by its cumulative methane emissions up to that year. Source per OWID codebook: Jones et al., National contributions to climate change (2025). | °C | measure |
| temperature_change_from_co2 | The entity's contribution to the change in global mean surface temperature since pre-industrial times caused by its cumulative CO2 emissions (fossil and land use) up to that year. Source per OWID codebook: Jones et al., National contributions to climate change (2025). | °C | measure |
| temperature_change_from_ghg | The entity's contribution to the change in global mean surface temperature since pre-industrial times caused by its cumulative emissions of CO2, CH4 and N2O combined up to that year. Source per OWID codebook: Jones et al., National contributions to climate change (2025). | °C | measure |
| temperature_change_from_n2o | The entity's contribution to the change in global mean surface temperature since pre-industrial times caused by its cumulative nitrous oxide emissions up to that year. Source per OWID codebook: Jones et al., National contributions to climate change (2025). | °C | measure |
| total_ghg | Annual total greenhouse-gas emissions (CO2 + CH4 + N2O) including land use, land-use change and forestry (LUCF), in CO2-equivalents (GWP100); can be negative where LUCF is a net sink. Source per OWID codebook: Jones et al., National contributions to climate change (2025). | million tonnes CO2-equivalent (GWP100) | measure |
| total_ghg_excluding_lucf | Annual total greenhouse-gas emissions (CO2 + CH4 + N2O) excluding land use, land-use change and forestry (LUCF), in CO2-equivalents (GWP100). Source per OWID codebook: Jones et al., National contributions to climate change (2025). | million tonnes CO2-equivalent (GWP100) | measure |
| trade_co2 | Net CO2 emissions embodied in trade = consumption_co2 minus co2 (verified exactly in the file). Positive = net importer of embodied emissions; negative = net exporter. Available 1990-2023 for about 120 countries. | million tonnes CO2 (Mt) | measure |
| trade_co2_share | Net CO2 embodied in trade as a percentage of territorial (production-based) CO2 emissions. | % | measure |
Caveats
- Aggregates and countries are mixed in the same rows: 34 aggregate or non-territorial entities without iso_code (World, continents, '(GCP)' regions, EU, income groups, OECD, International aviation/shipping, Kuwaiti Oil Fires) must be dropped before any cross-country analysis, or totals are double-counted. Kosovo and Ryukyu Islands also lack iso_code although they are not aggregates.
- The panel is highly unbalanced: every entity has a row for every year 1750-2024 (275 rows each), but most cells are empty before 1850-1950. Coverage differs by block: co2 1750-2024 (215 countries with data); population 1750-2024; land-use-change CO2, methane, N2O and total GHG 1850-2024 (about 199 countries); primary energy columns 1965-2024 (about 205); GDP 1820-2022 (164 countries, Maddison); consumption_co2 and trade_co2 only 1990-2023 for about 120 countries.
- GDP is in 2011 international dollars (PPP) from the Maddison Project; it ends in 2022, so any GDP-based measure (co2_per_gdp, energy_per_gdp, consumption_co2_per_gdp) is unavailable for 2023-2024. Maddison GDP for early years is reconstructed and uncertain.
- Many columns are deterministic transformations of others (per-capita = total / population; per-GDP = total*1e9 / gdp; growth_abs and growth_prct are first differences of the level; share_global_* = entity / World; cumulative_* = running sums; trade_co2 = consumption_co2 - co2, verified to within 0.0005 Mt). They carry no independent information and must not be used together as separate regressors.
- Extreme values reflect tiny denominators rather than errors: co2_growth_prct reaches 180,870% (from near-zero bases); co2_per_capita reaches 782.7 t/person and oil_co2_per_capita the same value, consistent with small oil-producing territories or very small populations; co2_including_luc_per_unit_energy ranges from -8,674 to 511,469 g/kWh.
- Land-use-change CO2 (and therefore co2_including_luc, total_ghg and ghg_per_capita) can be negative when land acts as a net sink, and is far more uncertain than fossil CO2 (bookkeeping-model estimates).
- Emissions series come from different sources and vintages: fossil and land-use CO2 from the Global Carbon Budget, CH4/N2O/total GHG and temperature contributions from Jones et al. (National contributions to climate change), energy from the EIA and Energy Institute; totals from different sources need not reconcile.
- Consumption-based emissions are model estimates from multi-regional input-output tables and are more uncertain than territorial emissions; for the World row consumption_co2 equals co2 by construction.
- The file was supplied by the user and has not been independently verified against the OWID release; the source names quoted above come from the public codebook, not from metadata inside the file.
Questions for you
- Please confirm the file is the unmodified Our World in Data CO2 dataset (owid-co2-data.csv) and from which release or download date, so the data can be cited correctly.
- Should analysis be restricted to the 218 entities with an ISO code (dropping all aggregates, International aviation/shipping and Kuwaiti Oil Fires)?
- Is the intended sample the period with GDP and consumption-based emissions both available (1990-2022), or should longer historical series be used?
- Should emissions be measured excluding land-use change (co2) as the main outcome, with land-use and total GHG measures treated as secondary?
Profile
Computed from the file itself, not by a model.
| Column | Type | Missing | Distinct | Summary |
|---|---|---|---|---|
| country | text | 0 | 254 | Afghanistan (275) · Africa (275) · Andorra (275) · Antarctica (275) · Asia (275) |
| year | integer | 0 | 275 | mean 1,920.35 · sd 65.8591 · median 1,925 · 1,750 to 2,024 |
| iso_code | text | 7,931 (15.7%) | 218 | AFG (275) · AND (275) · ATA (275) · AUS (275) · BHR (275) |
| population | integer | 9,244 (18.3%) | 40,197 | mean 60,172,900 · sd 330,841,000 · median 2,291,260 · 215 to 8,161,970,000 |
| gdp | integer | 35,160 (69.8%) | 15,249 | mean 330,049,000,000 · sd 3,086,380,000,000 · median 27,438,600,000 · 49,980,000 to 130,113,000,000,000 |
| cement_co2 | decimal | 21,238 (42.1%) | 10,944 | mean 7.89011 · sd 62.9882 · median 0 · 0 to 1,666.89 |
| cement_co2_per_capita | decimal | 24,763 (49.1%) | 13,185 | mean 0.0600262 · sd 0.123567 · median 0.00101756 · 0 to 2.4839 |
| co2 | decimal | 21,027 (41.7%) | 22,060 | mean 420.227 · sd 1,972.09 · median 5.08076 · 0 to 38,598.6 |
| co2_growth_abs | decimal | 23,195 (46%) | 19,649 | mean 6.26884 · sd 62.1995 · median 0.043968 · -1,928.34 to 1,804.66 |
| co2_growth_prct | decimal | 24,172 (47.9%) | 22,533 | mean 42.5982 · sd 1,721.91 · median 3.81348 · -100 to 180,870 |
| co2_including_luc | decimal | 26,615 (52.8%) | 23,320 | mean 544.145 · sd 2,273.28 · median 28.1198 · -84.5602 to 43,184.1 |
| co2_including_luc_growth_abs | decimal | 26,915 (53.4%) | 22,906 | mean 7.4837 · sd 99.5125 · median 0.111709 · -2,298.98 to 2,614.87 |
| co2_including_luc_growth_prct | decimal | 26,915 (53.4%) | 23,069 | mean 4.13234 · sd 166.964 · median 1.18332 · -11,927.2 to 6,362.5 |
| co2_including_luc_per_capita | decimal | 26,654 (52.9%) | 23,746 | mean 7.11545 · sd 8.92114 · median 4.53638 · -12.4546 to 364.809 |
| co2_including_luc_per_gdp | decimal | 33,621 (66.7%) | 16,771 | mean 2.32842 · sd 5.62688 · median 0.877673 · -12.9381 to 291.293 |
| co2_including_luc_per_unit_energy | decimal | 40,278 (79.9%) | 10,101 | mean 1,343.22 · sd 6,678.23 · median 282.23 · -8,673.52 to 511,469 |
| co2_per_capita | decimal | 23,902 (47.4%) | 26,024 | mean 3.82141 · sd 14.3129 · median 1.02337 · 0 to 782.743 |
| co2_per_gdp | decimal | 32,883 (65.2%) | 17,318 | mean 0.397679 · sd 0.754062 · median 0.263544 · 0 to 82.5989 |
| co2_per_unit_energy | decimal | 39,585 (78.5%) | 10,698 | mean 251.178 · sd 315.226 · median 220.868 · 0 to 10,688.9 |
| coal_co2 | decimal | 28,486 (56.5%) | 15,356 | mean 185.496 · sd 855.418 · median 2.09581 · 0 to 15,805.3 |
| coal_co2_per_capita | decimal | 29,097 (57.7%) | 19,619 | mean 1.28533 · sd 2.35109 · median 0.226842 · 0 to 34.2951 |
| consumption_co2 | decimal | 45,358 (90%) | 5,019 | mean 1,318.13 · sd 3,904.11 · median 74.3357 · 0 to 38,598.6 |
| consumption_co2_per_capita | decimal | 45,768 (90.8%) | 4,642 | mean 6.46363 · sd 6.8523 · median 4.32342 · 0 to 63.2716 |
| consumption_co2_per_gdp | decimal | 45,963 (91.2%) | 4,447 | mean 0.349924 · sd 0.247281 · median 0.301179 · 0 to 3.30573 |
| cumulative_cement_co2 | decimal | 21,260 (42.2%) | 12,741 | mean 196.486 · sd 1,495.92 · median 0 · 0 to 49,693.9 |
| cumulative_co2 | decimal | 22,848 (45.3%) | 23,920 | mean 12,492.2 · sd 73,121.4 · median 80.4665 · 0 to 1,849,120 |
| cumulative_co2_including_luc | decimal | 26,615 (52.8%) | 23,641 | mean 27,902.5 · sd 126,956 · median 1,039.96 · -120.345 to 2,751,500 |
| cumulative_coal_co2 | decimal | 28,486 (56.5%) | 19,139 | mean 9,311.02 · sd 43,500 · median 57.1547 · 0 to 849,832 |
| cumulative_flaring_co2 | decimal | 34,300 (68%) | 5,673 | mean 177.231 · sd 926.733 · median 0 · 0 to 20,141.4 |
| cumulative_gas_co2 | decimal | 32,264 (64%) | 9,431 | mean 1,876.82 · sd 11,756.9 · median 0.18934 · 0 to 276,573 |
| cumulative_luc_co2 | decimal | 12,961 (25.7%) | 35,363 | mean 9,998.38 · sd 46,043.1 · median 282.849 · -3,918.01 to 906,954 |
| cumulative_oil_co2 | decimal | 24,953 (49.5%) | 20,192 | mean 3,777.61 · sd 24,886.8 · median 24.131 · 0 to 638,014 |
| cumulative_other_co2 | decimal | 47,157 (93.6%) | 2,201 | mean 417.862 · sd 1,317.32 · median 21.4065 · 0 to 14,869.4 |
| energy_per_capita | decimal | 39,825 (79%) | 10,500 | mean 24,218 · sd 32,203.5 · median 11,793 · 0 to 318,560 |
| energy_per_gdp | decimal | 42,624 (84.6%) | 7,780 | mean 1.67291 · sd 1.62977 · median 1.24497 · 0 to 25.2525 |
| flaring_co2 | decimal | 34,237 (67.9%) | 5,366 | mean 5.85464 · sd 26.7499 · median 0 · 0 to 432.392 |
| flaring_co2_per_capita | decimal | 35,501 (70.4%) | 5,748 | mean 0.270544 · sd 2.99505 · median 0 · 0 to 114.509 |
| gas_co2 | decimal | 32,264 (64%) | 8,270 | mean 74.8404 · sd 397.297 · median 0.01832 · 0 to 8,009.83 |
| gas_co2_per_capita | decimal | 32,965 (65.4%) | 9,809 | mean 0.797033 · sd 2.80514 · median 0.00381143 · 0 to 53.4757 |
| ghg_excluding_lucf_per_capita | decimal | 14,569 (28.9%) | 35,830 | mean 3.07991 · sd 7.78545 · median 0.660426 · 0.0221263 to 369.65 |
| ghg_per_capita | decimal | 14,232 (28.2%) | 36,167 | mean 7.71684 · sd 11.2768 · median 4.27788 · -28.8521 to 369.705 |
| land_use_change_co2 | decimal | 12,961 (25.7%) | 32,760 | mean 124.753 · sd 519.819 · median 3.39649 · -319.418 to 8,692.77 |
| land_use_change_co2_per_capita | decimal | 13,776 (27.3%) | 35,080 | mean 3.90551 · sd 9.1066 · median 1.30575 · -45.3172 to 305.195 |
| methane | decimal | 12,261 (24.3%) | 38,142 | mean 95.2777 · sd 434.707 · median 3.46264 · 0.000149679 to 9,498.91 |
| methane_per_capita | decimal | 14,232 (28.2%) | 36,168 | mean 1.74147 · sd 3.8327 · median 1.00646 · 0.0318199 to 133.104 |
| nitrous_oxide | decimal | 11,911 (23.6%) | 37,983 | mean 25.4248 · sd 129.897 · median 0.706353 · 0 to 2,935.78 |
| nitrous_oxide_per_capita | decimal | 13,972 (27.7%) | 35,997 | mean 0.444592 · sd 0.799287 · median 0.224094 · 0 to 20.393 |
| oil_co2 | decimal | 24,952 (49.5%) | 14,443 | mean 115.308 · sd 646.557 · median 1.39847 · 0 to 12,470.6 |
| oil_co2_per_capita | decimal | 25,689 (51%) | 21,667 | mean 2.19272 · sd 13.9488 · median 0.404336 · 0 to 782.743 |
| other_co2_per_capita | decimal | 47,752 (94.7%) | 2,550 | mean 0.0822148 · sd 0.0797254 · median 0.0652546 · 0 to 0.604292 |
| other_industry_co2 | decimal | 47,157 (93.6%) | 2,085 | mean 20.1166 · sd 55.6447 · median 1.32138 · 0 to 458.178 |
| primary_energy_consumption | decimal | 39,780 (78.9%) | 10,267 | mean 2,450.93 · sd 11,383.3 · median 58.1856 · 0 to 176,737 |
| share_global_cement_co2 | decimal | 28,141 (55.8%) | 12,455 | mean 3.00281 · sd 13.39 · median 0.0264922 · 0 to 100 |
| share_global_co2 | decimal | 22,848 (45.3%) | 24,946 | mean 5.36987 · sd 18.7349 · median 0.0525134 · 0 to 100 |
| share_global_co2_including_luc | decimal | 26,615 (52.8%) | 23,523 | mean 3.45752 · sd 11.7097 · median 0.175536 · -0.374885 to 100 |
| share_global_coal_co2 | decimal | 28,486 (56.5%) | 18,559 | mean 6.79173 · sd 20.9695 · median 0.0579014 · 0 to 100 |
| share_global_cumulative_cement_co2 | decimal | 28,141 (55.8%) | 13,126 | mean 2.99566 · sd 13.4593 · median 0.0205496 · 0 to 100 |
| share_global_cumulative_co2 | decimal | 22,848 (45.3%) | 25,802 | mean 5.50066 · sd 19.5036 · median 0.0292623 · 0 to 100 |
| share_global_cumulative_co2_including_luc | decimal | 26,615 (52.8%) | 23,579 | mean 3.32196 · sd 11.6752 · median 0.143647 · -0.101453 to 100 |
| share_global_cumulative_coal_co2 | decimal | 28,486 (56.5%) | 20,127 | mean 6.97437 · sd 21.7972 · median 0.0397479 · 0 to 100 |
| share_global_cumulative_flaring_co2 | decimal | 39,322 (78%) | 6,029 | mean 3.04367 · sd 11.2399 · median 0.00406052 · 0 to 100 |
| share_global_cumulative_gas_co2 | decimal | 35,244 (69.9%) | 9,734 | mean 4.10454 · sd 17.156 · median 0.0143451 · 0 to 100 |
| share_global_cumulative_luc_co2 | decimal | 12,961 (25.7%) | 35,640 | mean 2.55937 · sd 9.7733 · median 0.104942 · -0.518435 to 100 |
| share_global_cumulative_oil_co2 | decimal | 26,578 (52.7%) | 21,654 | mean 3.3518 · sd 13.2605 · median 0.0571968 · 0 to 100 |
| share_global_cumulative_other_co2 | decimal | 48,241 (95.7%) | 2,001 | mean 7.05368 · sd 17.5541 · median 0.431603 · 0 to 100 |
| share_global_flaring_co2 | decimal | 39,322 (78%) | 5,580 | mean 3.12638 · sd 11.1478 · median 0.00104453 · 0 to 100 |
| share_global_gas_co2 | decimal | 35,244 (69.9%) | 9,295 | mean 4.19569 · sd 16.6836 · median 0.0258159 · 0 to 100 |
| share_global_luc_co2 | decimal | 12,961 (25.7%) | 35,394 | mean 2.43972 · sd 9.80082 · median 0.0716826 · -6.9658 to 100 |
| share_global_oil_co2 | decimal | 26,578 (52.7%) | 20,629 | mean 3.33334 · sd 13.0312 · median 0.0673228 · 0 to 100 |
| share_global_other_co2 | decimal | 48,241 (95.7%) | 1,999 | mean 7.19061 · sd 17.449 · median 0.588111 · 0 to 100 |
| share_of_temperature_change_from_ghg | decimal | 9,173 (18.2%) | 38,975 | mean 2.27225 · sd 9.28234 · median 0.0811026 · -0.82386 to 100 |
| temperature_change_from_ch4 | decimal | 12,131 (24.1%) | 37,930 | mean 0.00291262 · sd 0.0153531 · median 0.000060355 · -0.000866473 to 0.37654 |
| temperature_change_from_co2 | decimal | 9,173 (18.2%) | 37,311 | mean 0.00804818 · sd 0.0456799 · median 0.000115813 · -0.0000667666 to 1.21637 |
| temperature_change_from_ghg | decimal | 9,173 (18.2%) | 38,930 | mean 0.0112561 · sd 0.0628821 · median 0.000191549 · -0.000613182 to 1.67836 |
| temperature_change_from_n2o | decimal | 12,131 (24.1%) | 37,765 | mean 0.000543235 · sd 0.00304101 · median 0.0000106161 · 0 to 0.0854531 |
| total_ghg | decimal | 12,261 (24.3%) | 38,141 | mean 490.8 · sd 2,414.08 · median 14.6059 · -19.7252 to 54,433.4 |
| total_ghg_excluding_lucf | decimal | 12,598 (25%) | 37,810 | mean 310.521 · sd 1,812.36 · median 2.22192 · 0.0000733599 to 43,714.8 |
| trade_co2 | decimal | 45,699 (90.6%) | 4,575 | mean -6.98679 · sd 259.018 · median 1.64074 · -2,177.81 to 1,768.85 |
| trade_co2_share | decimal | 45,699 (90.6%) | 4,605 | mean 21.4686 · sd 62.6376 · median 8.3818 · -98.2811 to 1,023.04 |
First rows
{"country":"Afghanistan","year":1750,"iso_code":"AFG","population":2802560,"gdp":null,"cement_co2":0,"cement_co2_per_capita":0,"co2":null,"co2_growth_abs":null,"co2_growth_prct":null,"co2_including_luc":null,"co2_including_luc_growth_abs":null,"co2_including_luc_growth_prct":null,"co2_including_luc_per_capita":null,"co2_including_luc_per_gdp":null,"co2_including_luc_per_unit_energy":null,"co2_per_capita":null,"co2_per_gdp":null,"co2_per_unit_energy":null,"coal_co2":null,"coal_co2_per_capita":null,"consumption_co2":null,"consumption_co2_per_capita":null,"consumption_co2_per_gdp":null,"cumulative_cement_co2":0,"cumulative_co2":null,"cumulative_co2_including_luc":null,"cumulative_coal_co2":null,"cumulative_flaring_co2":null,"cumulative_gas_co2":null,"cumulative_luc_co2":null,"cumulative_oil_co2":null,"cumulative_other_co2":null,"energy_per_capita":null,"energy_per_gdp":null,"flaring_co2":null,"flaring_co2_per_capita":null,"gas_co2":null,"gas_co2_per_capita":null,"ghg_excluding_lucf_per_capita":null,"ghg_per_capita":null,"land_use_change_co2":null,"land_use_change_co2_per_capita":null,"methane":null,"methane_per_capita":null,"nitrous_oxide":null,"nitrous_oxide_per_capita":null,"oil_co2":null,"oil_co2_per_capita":null,"other_co2_per_capita":null,"other_industry_co2":null,"primary_energy_consumption":null,"share_global_cement_co2":null,"share_global_co2":null,"share_global_co2_including_luc":null,"share_global_coal_co2":null,"share_global_cumulative_cement_co2":null,"share_global_cumulative_co2":null,"share_global_cumulative_co2_including_luc":null,"share_global_cumulative_coal_co2":null,"share_global_cumulative_flaring_co2":null,"share_global_cumulative_gas_co2":null,"share_global_cumulative_luc_co2":null,"share_global_cumulative_oil_co2":null,"share_global_cumulative_other_co2":null,"share_global_flaring_co2":null,"share_global_gas_co2":null,"share_global_luc_co2":null,"share_global_oil_co2":null,"share_global_other_co2":null,"share_of_temperature_change_from_ghg":null,"temperature_change_from_ch4":null,"temperature_change_from_co2":null,"temperature_change_from_ghg":null,"temperature_change_from_n2o":null,"total_ghg":null,"total_ghg_excluding_lucf":null,"trade_co2":null,"trade_co2_share":null}
{"country":"Afghanistan","year":1751,"iso_code":"AFG","population":null,"gdp":null,"cement_co2":0,"cement_co2_per_capita":null,"co2":null,"co2_growth_abs":null,"co2_growth_prct":null,"co2_including_luc":null,"co2_including_luc_growth_abs":null,"co2_including_luc_growth_prct":null,"co2_including_luc_per_capita":null,"co2_including_luc_per_gdp":null,"co2_including_luc_per_unit_energy":null,"co2_per_capita":null,"co2_per_gdp":null,"co2_per_unit_energy":null,"coal_co2":null,"coal_co2_per_capita":null,"consumption_co2":null,"consumption_co2_per_capita":null,"consumption_co2_per_gdp":null,"cumulative_cement_co2":0,"cumulative_co2":null,"cumulative_co2_including_luc":null,"cumulative_coal_co2":null,"cumulative_flaring_co2":null,"cumulative_gas_co2":null,"cumulative_luc_co2":null,"cumulative_oil_co2":null,"cumulative_other_co2":null,"energy_per_capita":null,"energy_per_gdp":null,"flaring_co2":null,"flaring_co2_per_capita":null,"gas_co2":null,"gas_co2_per_capita":null,"ghg_excluding_lucf_per_capita":null,"ghg_per_capita":null,"land_use_change_co2":null,"land_use_change_co2_per_capita":null,"methane":null,"methane_per_capita":null,"nitrous_oxide":null,"nitrous_oxide_per_capita":null,"oil_co2":null,"oil_co2_per_capita":null,"other_co2_per_capita":null,"other_industry_co2":null,"primary_energy_consumption":null,"share_global_cement_co2":null,"share_global_co2":null,"share_global_co2_including_luc":null,"share_global_coal_co2":null,"share_global_cumulative_cement_co2":null,"share_global_cumulative_co2":null,"share_global_cumulative_co2_including_luc":null,"share_global_cumulative_coal_co2":null,"share_global_cumulative_flaring_co2":null,"share_global_cumulative_gas_co2":null,"share_global_cumulative_luc_co2":null,"share_global_cumulative_oil_co2":null,"share_global_cumulative_other_co2":null,"share_global_flaring_co2":null,"share_global_gas_co2":null,"share_global_luc_co2":null,"share_global_oil_co2":null,"share_global_other_co2":null,"share_of_temperature_change_from_ghg":null,"temperature_change_from_ch4":null,"temperature_change_from_co2":null,"temperature_change_from_ghg":null,"temperature_change_from_n2o":null,"total_ghg":null,"total_ghg_excluding_lucf":null,"trade_co2":null,"trade_co2_share":null}
{"country":"Afghanistan","year":1752,"iso_code":"AFG","population":null,"gdp":null,"cement_co2":0,"cement_co2_per_capita":null,"co2":null,"co2_growth_abs":null,"co2_growth_prct":null,"co2_including_luc":null,"co2_including_luc_growth_abs":null,"co2_including_luc_growth_prct":null,"co2_including_luc_per_capita":null,"co2_including_luc_per_gdp":null,"co2_including_luc_per_unit_energy":null,"co2_per_capita":null,"co2_per_gdp":null,"co2_per_unit_energy":null,"coal_co2":null,"coal_co2_per_capita":null,"consumption_co2":null,"consumption_co2_per_capita":null,"consumption_co2_per_gdp":null,"cumulative_cement_co2":0,"cumulative_co2":null,"cumulative_co2_including_luc":null,"cumulative_coal_co2":null,"cumulative_flaring_co2":null,"cumulative_gas_co2":null,"cumulative_luc_co2":null,"cumulative_oil_co2":null,"cumulative_other_co2":null,"energy_per_capita":null,"energy_per_gdp":null,"flaring_co2":null,"flaring_co2_per_capita":null,"gas_co2":null,"gas_co2_per_capita":null,"ghg_excluding_lucf_per_capita":null,"ghg_per_capita":null,"land_use_change_co2":null,"land_use_change_co2_per_capita":null,"methane":null,"methane_per_capita":null,"nitrous_oxide":null,"nitrous_oxide_per_capita":null,"oil_co2":null,"oil_co2_per_capita":null,"other_co2_per_capita":null,"other_industry_co2":null,"primary_energy_consumption":null,"share_global_cement_co2":null,"share_global_co2":null,"share_global_co2_including_luc":null,"share_global_coal_co2":null,"share_global_cumulative_cement_co2":null,"share_global_cumulative_co2":null,"share_global_cumulative_co2_including_luc":null,"share_global_cumulative_coal_co2":null,"share_global_cumulative_flaring_co2":null,"share_global_cumulative_gas_co2":null,"share_global_cumulative_luc_co2":null,"share_global_cumulative_oil_co2":null,"share_global_cumulative_other_co2":null,"share_global_flaring_co2":null,"share_global_gas_co2":null,"share_global_luc_co2":null,"share_global_oil_co2":null,"share_global_other_co2":null,"share_of_temperature_change_from_ghg":null,"temperature_change_from_ch4":null,"temperature_change_from_co2":null,"temperature_change_from_ghg":null,"temperature_change_from_n2o":null,"total_ghg":null,"total_ghg_excluding_lucf":null,"trade_co2":null,"trade_co2_share":null}
{"country":"Afghanistan","year":1753,"iso_code":"AFG","population":null,"gdp":null,"cement_co2":0,"cement_co2_per_capita":null,"co2":null,"co2_growth_abs":null,"co2_growth_prct":null,"co2_including_luc":null,"co2_including_luc_growth_abs":null,"co2_including_luc_growth_prct":null,"co2_including_luc_per_capita":null,"co2_including_luc_per_gdp":null,"co2_including_luc_per_unit_energy":null,"co2_per_capita":null,"co2_per_gdp":null,"co2_per_unit_energy":null,"coal_co2":null,"coal_co2_per_capita":null,"consumption_co2":null,"consumption_co2_per_capita":null,"consumption_co2_per_gdp":null,"cumulative_cement_co2":0,"cumulative_co2":null,"cumulative_co2_including_luc":null,"cumulative_coal_co2":null,"cumulative_flaring_co2":null,"cumulative_gas_co2":null,"cumulative_luc_co2":null,"cumulative_oil_co2":null,"cumulative_other_co2":null,"energy_per_capita":null,"energy_per_gdp":null,"flaring_co2":null,"flaring_co2_per_capita":null,"gas_co2":null,"gas_co2_per_capita":null,"ghg_excluding_lucf_per_capita":null,"ghg_per_capita":null,"land_use_change_co2":null,"land_use_change_co2_per_capita":null,"methane":null,"methane_per_capita":null,"nitrous_oxide":null,"nitrous_oxide_per_capita":null,"oil_co2":null,"oil_co2_per_capita":null,"other_co2_per_capita":null,"other_industry_co2":null,"primary_energy_consumption":null,"share_global_cement_co2":null,"share_global_co2":null,"share_global_co2_including_luc":null,"share_global_coal_co2":null,"share_global_cumulative_cement_co2":null,"share_global_cumulative_co2":null,"share_global_cumulative_co2_including_luc":null,"share_global_cumulative_coal_co2":null,"share_global_cumulative_flaring_co2":null,"share_global_cumulative_gas_co2":null,"share_global_cumulative_luc_co2":null,"share_global_cumulative_oil_co2":null,"share_global_cumulative_other_co2":null,"share_global_flaring_co2":null,"share_global_gas_co2":null,"share_global_luc_co2":null,"share_global_oil_co2":null,"share_global_other_co2":null,"share_of_temperature_change_from_ghg":null,"temperature_change_from_ch4":null,"temperature_change_from_co2":null,"temperature_change_from_ghg":null,"temperature_change_from_n2o":null,"total_ghg":null,"total_ghg_excluding_lucf":null,"trade_co2":null,"trade_co2_share":null}
{"country":"Afghanistan","year":1754,"iso_code":"AFG","population":null,"gdp":null,"cement_co2":0,"cement_co2_per_capita":null,"co2":null,"co2_growth_abs":null,"co2_growth_prct":null,"co2_including_luc":null,"co2_including_luc_growth_abs":null,"co2_including_luc_growth_prct":null,"co2_including_luc_per_capita":null,"co2_including_luc_per_gdp":null,"co2_including_luc_per_unit_energy":null,"co2_per_capita":null,"co2_per_gdp":null,"co2_per_unit_energy":null,"coal_co2":null,"coal_co2_per_capita":null,"consumption_co2":null,"consumption_co2_per_capita":null,"consumption_co2_per_gdp":null,"cumulative_cement_co2":0,"cumulative_co2":null,"cumulative_co2_including_luc":null,"cumulative_coal_co2":null,"cumulative_flaring_co2":null,"cumulative_gas_co2":null,"cumulative_luc_co2":null,"cumulative_oil_co2":null,"cumulative_other_co2":null,"energy_per_capita":null,"energy_per_gdp":null,"flaring_co2":null,"flaring_co2_per_capita":null,"gas_co2":null,"gas_co2_per_capita":null,"ghg_excluding_lucf_per_capita":null,"ghg_per_capita":null,"land_use_change_co2":null,"land_use_change_co2_per_capita":null,"methane":null,"methane_per_capita":null,"nitrous_oxide":null,"nitrous_oxide_per_capita":null,"oil_co2":null,"oil_co2_per_capita":null,"other_co2_per_capita":null,"other_industry_co2":null,"primary_energy_consumption":null,"share_global_cement_co2":null,"share_global_co2":null,"share_global_co2_including_luc":null,"share_global_coal_co2":null,"share_global_cumulative_cement_co2":null,"share_global_cumulative_co2":null,"share_global_cumulative_co2_including_luc":null,"share_global_cumulative_coal_co2":null,"share_global_cumulative_flaring_co2":null,"share_global_cumulative_gas_co2":null,"share_global_cumulative_luc_co2":null,"share_global_cumulative_oil_co2":null,"share_global_cumulative_other_co2":null,"share_global_flaring_co2":null,"share_global_gas_co2":null,"share_global_luc_co2":null,"share_global_oil_co2":null,"share_global_other_co2":null,"share_of_temperature_change_from_ghg":null,"temperature_change_from_ch4":null,"temperature_change_from_co2":null,"temperature_change_from_ghg":null,"temperature_change_from_n2o":null,"total_ghg":null,"total_ghg_excluding_lucf":null,"trade_co2":null,"trade_co2_share":null}2.Literature review
Literature review complete (live:agent:claude-opus-5-5): 43 key papers, 43 grounded sources, 5 open gaps.
Summary
The income-emissions literature runs from the EKC (Grossman and Krueger 1995; Holtz-Eakin and Selden 1995; Schmalensee et al. 1998) through its econometric critiques (Stern 2004, 2017; Wagner 2008) to a recent decoupling literature that counts episodes of absolute decoupling (Le Quere et al. 2019; Hubacek et al. 2021; Vogel and Hickel 2023; Haberl et al. 2020). A parallel line builds consumption-based (trade-adjusted) inventories (Davis and Caldeira 2010; Peters et al. 2011, 2012) and finds that rich countries' territorial stabilisation was partly offset by rising emissions embodied in imports until net transfers peaked in the mid-2000s (Wood et al. 2020). Panel comparisons of GDP elasticities across the two accounting bases (Knight and Schor 2014; Fernandez-Amador et al. 2017; Cohen et al. 2018) suggest consumption-based emissions are more GDP-elastic in rich countries, but they use small samples ending by 2008-2011. Economics evidence on leakage is mixed: Kyoto commitments raised embodied imports (Aichele and Felbermayr 2012, 2015), whereas the EU ETS shows no leakage (Naegele and Zaklan 2019; Dechezlepretre et al. 2022), and US manufacturing cleanup was mainly technique (Levinson 2009; Shapiro and Walker 2018). Whether the consumption-territorial elasticity gap is general across income levels, and whether it closed after emission transfers peaked, is untested on the full 1990-2022 panel.
Key papers (43)
- Economic Growth and the Environment. Grossman and Krueger; Quarterly Journal of Economics, 1995.Founding EKC paper: inverted-U between income and local pollutants; the benchmark for income-emissions panels.
- Stoking the fires? CO2 emissions and economic growth. Holtz-Eakin and Selden; Journal of Public Economics, 1995.Within-country CO2-income panel; turning point far outside sample, so growth raises CO2.
- World Carbon Dioxide Emissions: 1950-2050. Schmalensee, Stoker and Judson; Review of Economics and Statistics, 1998.Flexible fixed-effects CO2-income panel of exactly this country-year type.
- Confronting the Environmental Kuznets Curve. Dasgupta, Laplante, Wang and Wheeler; Journal of Economic Perspectives, 2002.Review of EKC evidence and its fragility.
- Trade, Growth, and the Environment. Copeland and Taylor; Journal of Economic Literature, 2004.Scale/composition/technique decomposition and pollution-haven logic linking trade and emissions.
- The Rise and Fall of the Environmental Kuznets Curve. Stern; World Development, 2004.Econometric critique of EKC estimates.
- The environmental Kuznets curve after 25 years. Stern; Journal of Bioeconomics, 2017.Updated review; favours growth-rate and long-difference approaches.
- The carbon Kuznets curve: A cloudy picture emitted by bad econometrics?. Wagner; Resource and Energy Economics, 2008.Nonstationarity and cross-sectional dependence undo carbon-EKC findings; motivates differenced specifications and dependence-robust inference.
- Global and regional drivers of accelerating CO2 emissions. Raupach et al.; PNAS, 2007.Kaya-identity decomposition of emission growth.
- Technology, International Trade, and Pollution from US Manufacturing. Levinson; American Economic Review, 2009.Decline in US manufacturing pollution driven by technique, not offshoring: a benchmark against displacement.
- Consumption-based accounting of CO2 emissions. Davis and Caldeira; PNAS, 2010.MRIO consumption-based inventories; magnitude of emissions embodied in trade.
- Growth in emission transfers via international trade from 1990 to 2008. Peters, Minx, Weber and Edenhofer; PNAS, 2011.Net transfers to developed countries rose from 0.4 to 1.6 Gt CO2 1990-2008; consumption emissions grew faster than territorial in most developed countries.
- Kyoto and the carbon footprint of nations. Aichele and Felbermayr; Journal of Environmental Economics and Management, 2012.Causal evidence that Kyoto commitment raised footprints relative to territorial emissions.
- A synthesis of carbon in international trade. Peters, Davis and Andrew; Biogeosciences, 2012.Synthesis of consumption-based accounting methods behind GCB consumption series.
- Evidence on business cycles and CO2 emissions. Doda; Journal of Macroeconomics, 2014.Cyclical CO2 procyclical; cyclical elasticity varies with income.
- Economic Growth and Climate Change: A Cross-National Analysis of Territorial and Consumption-Based Carbon Emissions in High-Income Countries. Knight and Schor; Sustainability, 2014.Closest prior: GDP effect larger for consumption-based than territorial emissions in 29 rich countries, 1991-2008.
- Kyoto and Carbon Leakage: An Empirical Analysis of the Carbon Content of Bilateral Trade. Aichele and Felbermayr; Review of Economics and Statistics, 2015.Gravity estimate: binding Kyoto commitments raised embodied carbon imports from non-committed countries by about 8%.
- Carbon dioxide emissions in the short run: The rate and sources of economic growth matter. Burke, Shahiduzzaman and Stern; Global Environmental Change, 2015.Short-run CO2-GDP elasticities depend on the rate and sources of growth.
- National greenhouse-gas accounting for effective climate policy on international trade. Kander, Jiborn, Moran and Wiedmann; Nature Climate Change, 2015.Technology-adjusted consumption accounting; conventional measure may overstate displacement.
- Carbon Dioxide Emissions and Economic Growth: An Assessment Based on Production and Consumption Emission Inventories. Fernandez-Amador, Francois, Oberdabernig and Tomberger; Ecological Economics, 2017.Income elasticities of production and consumption inventories 1997-2011; regime-dependent, small efficiency gains.
- Asymmetric effects of the business cycle on carbon dioxide emissions. Sheldon; Energy Economics, 2017.Asymmetric expansion/recession elasticities of CO2.
- Trade and the Environment: New Methods, Measurements, and Results. Cherniwchan, Copeland and Taylor; Annual Review of Economics, 2017.Survey of trade-environment identification methods.
- Decoupling or delusion? Measuring emissions displacement in foreign trade. Jiborn, Kander, Kulionis, Nielsen and Moran; Global Environmental Change, 2018.Displacement smaller once technology differences adjusted (Sweden, UK 1995-2009).
- The long-run decoupling of emissions and output: Evidence from the largest emitters. Cohen, Jalles, Loungani and Marto; Energy Policy, 2018.Trend decoupling in rich countries weakens on a consumption basis; link weakened in recent decades.
- Why Is Pollution from US Manufacturing Declining? The Roles of Environmental Regulation, Productivity, and Trade. Shapiro and Walker; American Economic Review, 2018.Regulation-driven technique effects dominate trade in explaining pollution declines.
- Trends of the EU's territorial and consumption-based emissions from 1990 to 2016. Karstensen, Peters and Andrew; Climatic Change, 2018.EU territorial vs consumption trajectories and the role of trade.
- Drivers of declining CO2 emissions in 18 developed economies. Le Quere et al.; Nature Climate Change, 2019.Absolute declines 2005-2015 in 18 developed economies; energy and renewables drivers.
- Climate Change: The Ultimate Challenge for Economics. Nordhaus; American Economic Review, 2019.Macro-climate-economics framing.
- Does the EU ETS cause carbon leakage in European manufacturing?. Naegele and Zaklan; Journal of Environmental Economics and Management, 2019.No evidence of ETS-induced leakage.
- Beyond peak emission transfers: historical impacts of globalization and future impacts of climate policies on international emission transfers. Wood et al.; Climate Policy, 2020.Net emission transfers peaked mid-2000s: motivates pre/post-2007 comparison.
- A systematic review of the evidence on decoupling of GDP, resource use and GHG emissions, part II: synthesizing the insights. Haberl et al.; Environmental Research Letters, 2020.Absolute decoupling rare and slow.
- The Environmental Bias of Trade Policy. Shapiro; Quarterly Journal of Economics, 2021.Trade policy implicitly subsidises CO2-intensive goods.
- Evidence of decoupling consumption-based CO2 emissions from economic growth. Hubacek, Chen, Feng, Wiedmann and Shan; Advances in Applied Energy, 2021.32 of 116 countries decoupled production-based and 23 consumption-based emissions in 2015-2018.
- A review of trends and drivers of greenhouse gas emissions by sector from 1990 to 2018. Lamb et al.; Environmental Research Letters, 2021.Sectoral trends and drivers since 1990.
- Searching for carbon leaks in multinational companies. Dechezlepretre, Gennaioli, Martin, Muuls and Stoerk; Journal of Environmental Economics and Management, 2022.No leakage within multinationals.
- Carbon Leakage, Consumption, and Trade. Grubb et al.; Annual Review of Environment and Resources, 2022.Review of leakage and consumption-based evidence.
- Is green growth happening? An empirical analysis of achieved versus Paris-compliant CO2-GDP decoupling in high-income countries. Vogel and Hickel; Lancet Planetary Health, 2023.11 rich countries decoupled consumption CO2 2013-2019 at rates far below Paris-compliant.
- Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data. Driscoll and Kraay; Review of Economics and Statistics, 1998.Standard errors robust to cross-sectional and serial dependence in country-year panels.
- A simple panel unit root test in the presence of cross-section dependence. Pesaran; Journal of Applied Econometrics, 2007.CIPS unit-root test under cross-sectional dependence.
- National contributions to climate change due to historical emissions of carbon dioxide, methane, and nitrous oxide since 1850. Jones et al.; Scientific Data, 2023.Source of CH4, N2O, total GHG and temperature-contribution columns.
- Global Carbon Budget 2024. Friedlingstein et al.; Earth System Science Data, 2025.Global Carbon Budget methodology for fossil, land-use and consumption-based CO2.
- Global Carbon Budget 2025. Friedlingstein et al.; Earth System Science Data, 2026.Release cited by the OWID codebook as the source of the CO2 columns.
- Maddison-style estimates of the evolution of the world economy: A new 2023 update. Bolt and van Zanden; Journal of Economic Surveys, 2025.Source of the GDP column (Maddison Project Database 2023).
Methods in use
- Log-log panel regressions of emissions on GDP with country and year fixed effects (within-country elasticity)
- First-difference / annual growth-rate regressions and multi-year long differences to avoid spurious regression with integrated series
- EKC polynomial and threshold specifications in log income
- Trend/cycle decompositions separating long-run from cyclical elasticities
- Heterogeneity by income group or Kyoto status via interactions; sub-period splits around the mid-2000s
- Driscoll-Kraay and country-clustered standard errors; Pesaran cross-sectional dependence and CIPS unit-root tests
- MRIO consumption-based accounting and balances of emissions embodied in trade; technology-adjusted balances
- Gravity models of the carbon content of bilateral trade with policy-treatment variables
- Counts of absolute/relative decoupling episodes over fixed windows
Open gaps
- Production- versus consumption-based GDP elasticities have been compared only on small samples ending 2008-2011; the full 1990-2022 Global Carbon Budget consumption panel (about 120 countries) has not been used for a formal test with dependence-robust inference.
- Whether the elasticity gap (displacement wedge) varies systematically with income level - positive for rich net importers, negative for poor net exporters - is untested.
- Whether the wedge closed after net emission transfers peaked in the mid-2000s has not been tested formally.
- Long-run (long-difference) versus short-run (annual) elasticities across the two accounting bases are rarely contrasted.
- Whether territorial absolute decoupling survives consumption accounting over 2007-2022 for the same set of countries.
Grounded search sources (43) · mode: live:agent:claude-opus-5-5
- Economic Growth and the Environment (Grossman and Krueger, QJE 1995)
- Stoking the fires? CO2 emissions and economic growth (Holtz-Eakin and Selden, Journal of Public Economics 1995)
- World Carbon Dioxide Emissions: 1950-2050 (Schmalensee, Stoker and Judson, REStat 1998)
- Confronting the Environmental Kuznets Curve (Dasgupta, Laplante, Wang and Wheeler, JEP 2002)
- Trade, Growth, and the Environment (Copeland and Taylor, JEL 2004)
- The Rise and Fall of the Environmental Kuznets Curve (Stern, World Development 2004)
- The environmental Kuznets curve after 25 years (Stern, Journal of Bioeconomics 2017)
- The carbon Kuznets curve: A cloudy picture emitted by bad econometrics? (Wagner, Resource and Energy Economics 2008)
- Global and regional drivers of accelerating CO2 emissions (Raupach et al., PNAS 2007)
- Technology, International Trade, and Pollution from US Manufacturing (Levinson, AER 2009)
- Consumption-based accounting of CO2 emissions (Davis and Caldeira, PNAS 2010)
- Growth in emission transfers via international trade from 1990 to 2008 (Peters, Minx, Weber and Edenhofer, PNAS 2011)
- Kyoto and the carbon footprint of nations (Aichele and Felbermayr, JEEM 2012)
- A synthesis of carbon in international trade (Peters, Davis and Andrew, Biogeosciences 2012)
- Evidence on business cycles and CO2 emissions (Doda, Journal of Macroeconomics 2014)
- Economic Growth and Climate Change: A Cross-National Analysis of Territorial and Consumption-Based Carbon Emissions in High-Income Countries (Knight and Schor, Sustainability 2014)
- Kyoto and Carbon Leakage: An Empirical Analysis of the Carbon Content of Bilateral Trade (Aichele and Felbermayr, REStat 2015)
- Carbon dioxide emissions in the short run: The rate and sources of economic growth matter (Burke, Shahiduzzaman and Stern, Global Environmental Change 2015)
- National greenhouse-gas accounting for effective climate policy on international trade (Kander, Jiborn, Moran and Wiedmann, Nature Climate Change 2015)
- Carbon Dioxide Emissions and Economic Growth: An Assessment Based on Production and Consumption Emission Inventories (Fernandez-Amador, Francois, Oberdabernig and Tomberger, Ecological Economics 2017)
- Asymmetric effects of the business cycle on carbon dioxide emissions (Sheldon, Energy Economics 2017)
- Trade and the Environment: New Methods, Measurements, and Results (Cherniwchan, Copeland and Taylor, Annual Review of Economics 2017)
- Decoupling or delusion? Measuring emissions displacement in foreign trade (Jiborn, Kander, Kulionis, Nielsen and Moran, Global Environmental Change 2018)
- The long-run decoupling of emissions and output: Evidence from the largest emitters (Cohen, Jalles, Loungani and Marto, Energy Policy 2018)
- Why Is Pollution from US Manufacturing Declining? The Roles of Environmental Regulation, Productivity, and Trade (Shapiro and Walker, AER 2018)
- Trends of the EU's territorial and consumption-based emissions from 1990 to 2016 (Karstensen, Peters and Andrew, Climatic Change 2018)
- Drivers of declining CO2 emissions in 18 developed economies (Le Quere et al., Nature Climate Change 2019)
- Climate Change: The Ultimate Challenge for Economics (Nordhaus, AER 2019)
- Does the EU ETS cause carbon leakage in European manufacturing? (Naegele and Zaklan, JEEM 2019)
- Beyond peak emission transfers: historical impacts of globalization and future impacts of climate policies on international emission transfers (Wood et al., Climate Policy 2020)
- A systematic review of the evidence on decoupling of GDP, resource use and GHG emissions, part II: synthesizing the insights (Haberl et al., Environmental Research Letters 2020)
- The Environmental Bias of Trade Policy (Shapiro, QJE 2021)
- Evidence of decoupling consumption-based CO2 emissions from economic growth (Hubacek, Chen, Feng, Wiedmann and Shan, Advances in Applied Energy 2021)
- A review of trends and drivers of greenhouse gas emissions by sector from 1990 to 2018 (Lamb et al., Environmental Research Letters 2021)
- Searching for carbon leaks in multinational companies (Dechezlepretre, Gennaioli, Martin, Muuls and Stoerk, JEEM 2022)
- Carbon Leakage, Consumption, and Trade (Grubb et al., Annual Review of Environment and Resources 2022)
- Is green growth happening? An empirical analysis of achieved versus Paris-compliant CO2-GDP decoupling in high-income countries (Vogel and Hickel, Lancet Planetary Health 2023)
- Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data (Driscoll and Kraay, REStat 1998)
- A simple panel unit root test in the presence of cross-section dependence (Pesaran, Journal of Applied Econometrics 2007)
- National contributions to climate change due to historical emissions of carbon dioxide, methane, and nitrous oxide since 1850 (Jones et al., Scientific Data 2023)
- Global Carbon Budget 2024 (Friedlingstein et al., Earth System Science Data 2025)
- Global Carbon Budget 2025 (Friedlingstein et al., Earth System Science Data 2026)
- Maddison-style estimates of the evolution of the world economy: A new 2023 update (Bolt and van Zanden, Journal of Economic Surveys)
3.Ideation
3 candidate ideas generated; selected: “Decoupling or displacement? The GDP elasticity of consumption-based versus territorial CO2 emissions, 1990-2022”.
- Idea 1Selected
Decoupling or displacement? The GDP elasticity of consumption-based versus territorial CO2 emissions, 1990-2022
Hypothesis. Among countries in the top tercile of 1990 real GDP per capita (computed from gdp/population), the within-country GDP elasticity of consumption-based CO2 (consumption_co2) exceeds that of territorial CO2 (co2) over 1990-2007, i.e. the 'displacement wedge' elasticity d log(consumption_co2/co2) / d log gdp is positive; and this wedge is smaller in 2007-2022, after net emission transfers peaked. Refuted if the rich-tercile wedge elasticity in 1990-2007 is not positive at the 5 percent level, or if the 2007-2022 wedge is not smaller than the 1990-2007 wedge.
Dataset. Your data
Columns used. country, iso_code, year, gdp, population, co2, consumption_co2, trade_co2, trade_co2_share, co2_per_capita, consumption_co2_per_capita
Test plan, novelty and risk
Test plan. Sample: entities with non-missing iso_code (countries/territories only; all aggregates dropped), years 1990-2022 (gdp ends 2022, consumption_co2 starts 1990), rows with gdp, population, co2>0 and consumption_co2>0; Kuwait 1990-1992 excluded ex ante because its co2 in 1991 includes the Gulf War oil fires (the file carries a separate 'Kuwaiti Oil Fires' entity). Income terciles from 1990 gdp/population (fixed at baseline, so not affected by later growth). (1) Primary - long differences: for each country, Delta log of co2, consumption_co2 and gdp over 1990-2007 and 2007-2022; regress Delta log co2 and Delta log consumption_co2 on Delta log gdp by tercile, and regress the wedge Delta log(consumption_co2/co2) on Delta log gdp interacted with tercile dummies; heteroskedasticity-robust (HC1) SEs; stacked two-period version with country-clustered SEs to test the change in the wedge between periods. (2) Annual first differences: Delta log E_it = beta * Delta log gdp_it + year FE, separately for co2 and consumption_co2 and for the wedge, interacted with tercile and a post-2007 dummy; Driscoll-Kraay SEs (robust to cross-sectional and serial dependence) and country-clustered SEs. (3) Accounting check: trade_co2 equals consumption_co2 - co2 in the file; report the rich-tercile mean trade_co2_share in 1990, 2007 and 2022. (4) Robustness: start in 1995 (post-transition), drop countries under 1 million people, per-capita versions (co2_per_capita, consumption_co2_per_capita on gdp/population), and a fixed-effects levels regression. Report point estimates, SEs in parentheses with clustering stated, stars at 10/5/1, N countries, N obs, R2.
Novelty. Knight and Schor (2014) and Cohen et al. (2018) compare the two accounting bases on 29 rich countries to 2008 or the 20 largest emitters; Fernandez-Amador et al. (2017) stop in 2011. None tests the wedge elasticity directly on the full ~120-country Global Carbon Budget consumption panel through 2022, none contrasts it across baseline-income terciles (where it should change sign between net importers and net exporters), and none tests formally whether it closed after the mid-2000s peak in emission transfers (Wood et al. 2020).
Expected outcome. A positive rich-tercile wedge in 1990-2007 (consumption-based emissions more GDP-elastic than territorial ones by a few tenths), a near-zero or negative wedge for the bottom tercile, and a smaller, statistically insignificant rich-tercile wedge in 2007-2022 - so that recent rich-country decoupling is not mainly displacement, while pre-2007 decoupling partly was.
Main risk. Elasticities are conditional associations, not causal effects of GDP; the consumption-based series is model-based (multi-regional input-output tables at benchmark years, interpolated between them), so annual changes in the wedge are partly constructed - hence long differences are primary. Only about 39 countries per tercile, so power is limited; Maddison GDP and Global Carbon Budget vintages may be revised.
- Idea 2
Energy intensity or carbon intensity? Decomposing the decline in the emissions elasticity of growth, 1965-2022
Hypothesis. The within-country elasticity of territorial CO2 with respect to GDP fell between 1965-1990 and 1990-2022, and in the top income tercile the fall is accounted for mainly by a lower elasticity of primary energy use to GDP rather than by a fall in the carbon intensity of energy. Refuted if the energy-use elasticity is unchanged across periods while the carbon-intensity component (co2_per_unit_energy) accounts for most of the change.
Dataset. Your data
Columns used. country, iso_code, year, gdp, population, co2, primary_energy_consumption, co2_per_unit_energy, energy_per_gdp
Test plan, novelty and risk
Test plan. Countries with iso_code, 1965-2022, rows with gdp, population, co2>0 and primary_energy_consumption>0. Kaya-style identity: log co2 = log primary_energy_consumption + log(co2/primary_energy_consumption); because the identity is exact, the GDP elasticity of co2 equals the sum of the GDP elasticities of energy and of carbon intensity (co2_per_unit_energy). Estimate each with annual first differences and year FE, interacted with period (1965-1990 vs 1991-2022) and baseline-income tercile; Driscoll-Kraay and country-clustered SEs; long-difference versions per period. Report the share of the change in the co2 elasticity attributable to each component with delta-method SEs.
Novelty. Le Quere et al. (2019) attribute 2005-2015 declines in 18 developed economies to energy and renewables descriptively; Raupach et al. (2007) decompose levels. A formal panel decomposition of the elasticity of growth, with dependence-robust inference across 60 years and income groups, is rarer.
Expected outcome. Elasticity of co2 to GDP falls across periods; in rich countries both components contribute, with energy intensity the larger share.
Main risk. Energy data start in 1965 and have changing coverage; the decomposition is mechanical, so it describes channels rather than causes; the oil shocks of the 1970s load on the early period.
- Idea 3
Asymmetric emissions responses to expansions and contractions across accounting bases
Hypothesis. Territorial CO2 responds more strongly to GDP contractions than to expansions (Sheldon 2017's asymmetry generalises to the global panel), and the asymmetry is weaker for consumption-based CO2. Refuted if the contraction and expansion elasticities are equal at the 5 percent level in the territorial series.
Dataset. Your data
Columns used. country, iso_code, year, gdp, co2, consumption_co2
Test plan, novelty and risk
Test plan. Countries with iso_code, 1990-2022, rows with gdp, co2>0, consumption_co2>0. Annual first differences: Delta log E = b+ * max(Delta log gdp, 0) + b- * min(Delta log gdp, 0) + year FE (+ country FE in robustness), separately for co2 and consumption_co2; Wald test b+ = b-; Driscoll-Kraay SEs. Robustness excluding 2009 and 2020 and excluding transition economies 1990-1995.
Novelty. Asymmetry has been studied for the US (Sheldon 2017) and with production-based data (Doda 2014; Burke et al. 2015); comparing it across consumption and territorial accounts is new.
Expected outcome. Somewhat larger contraction elasticity for territorial emissions; little difference for consumption-based emissions.
Main risk. Contractions are few and concentrated in 2009, 2020 and the early-1990s transition; consumption-based annual changes are partly interpolated.
Ranking rationale
Idea 1 answers the central policy question the decoupling literature leaves open - whether rich-country decoupling is real or displaced through trade - with a direct test (the wedge elasticity) that differences out shocks common to both accounts in the same country-year, uses the full 1990-2022 consumption panel, and has a sharp falsifiable prediction in two dimensions (income tercile, pre/post 2007). Idea 2 is credible but largely mechanical. Idea 3 rests on few contraction episodes and interpolated consumption data, so it is the weakest.
4.Experimentation
Experiment succeeded in 1 round (6.978s).
Data provenance
owid-co2-data.csv
- Format
- csv
- Observations
- 50,411
- Variables (79)
- country, year, iso_code, population, gdp, cement_co2, cement_co2_per_capita, co2, co2_growth_abs, co2_growth_prct, co2_including_luc, co2_including_luc_growth_abs, co2_including_luc_growth_prct, co2_including_luc_per_capita, co2_including_luc_per_gdp, co2_including_luc_per_unit_energy, co2_per_capita, co2_per_gdp, co2_per_unit_energy, coal_co2, coal_co2_per_capita, consumption_co2, consumption_co2_per_capita, consumption_co2_per_gdp, cumulative_cement_co2, cumulative_co2, cumulative_co2_including_luc, cumulative_coal_co2, cumulative_flaring_co2, cumulative_gas_co2, cumulative_luc_co2, cumulative_oil_co2, cumulative_other_co2, energy_per_capita, energy_per_gdp, flaring_co2, flaring_co2_per_capita, gas_co2, gas_co2_per_capita, ghg_excluding_lucf_per_capita, ghg_per_capita, land_use_change_co2, land_use_change_co2_per_capita, methane, methane_per_capita, nitrous_oxide, nitrous_oxide_per_capita, oil_co2, oil_co2_per_capita, other_co2_per_capita, other_industry_co2, primary_energy_consumption, share_global_cement_co2, share_global_co2, share_global_co2_including_luc, share_global_coal_co2, share_global_cumulative_cement_co2, share_global_cumulative_co2, share_global_cumulative_co2_including_luc, share_global_cumulative_coal_co2, share_global_cumulative_flaring_co2, share_global_cumulative_gas_co2, share_global_cumulative_luc_co2, share_global_cumulative_oil_co2, share_global_cumulative_other_co2, share_global_flaring_co2, share_global_gas_co2, share_global_luc_co2, share_global_oil_co2, share_global_other_co2, share_of_temperature_change_from_ghg, temperature_change_from_ch4, temperature_change_from_co2, temperature_change_from_ghg, temperature_change_from_n2o, total_ghg, total_ghg_excluding_lucf, trade_co2, trade_co2_share
- Units
- country: none; year: year; iso_code: none; population: people; gdp: international-$ at 2011 prices; cement_co2: million tonnes (Mt); cement_co2_per_capita: tonnes per person (t/person); co2: million tonnes CO2 (Mt); co2_growth_abs: million tonnes (Mt); co2_growth_prct: %; co2_including_luc: million tonnes (Mt); co2_including_luc_growth_abs: million tonnes (Mt); co2_including_luc_growth_prct: %; co2_including_luc_per_capita: tonnes per person (t/person); co2_including_luc_per_gdp: kilograms per international-$ (kg/$); co2_including_luc_per_unit_energy: grams per kilowatt-hour (g/kWh); co2_per_capita: tonnes per person (t/person); co2_per_gdp: kg CO2 per international-$ (2011 prices); co2_per_unit_energy: grams per kilowatt-hour (g/kWh); coal_co2: million tonnes (Mt); coal_co2_per_capita: tonnes per person (t/person); consumption_co2: million tonnes CO2 (Mt); consumption_co2_per_capita: tonnes per person (t/person); consumption_co2_per_gdp: kilograms per international-$ (kg/$); cumulative_cement_co2: million tonnes (Mt); cumulative_co2: million tonnes (Mt); cumulative_co2_including_luc: million tonnes (Mt); cumulative_coal_co2: million tonnes (Mt); cumulative_flaring_co2: million tonnes (Mt); cumulative_gas_co2: million tonnes (Mt); cumulative_luc_co2: million tonnes (Mt); cumulative_oil_co2: million tonnes (Mt); cumulative_other_co2: million tonnes (Mt); energy_per_capita: kWh per person; energy_per_gdp: kWh per international-$ (2011 prices); flaring_co2: million tonnes (Mt); flaring_co2_per_capita: tonnes per person (t/person); gas_co2: million tonnes (Mt); gas_co2_per_capita: tonnes per person (t/person); ghg_excluding_lucf_per_capita: t CO2-equivalent per person; ghg_per_capita: t CO2-equivalent per person; land_use_change_co2: million tonnes (Mt); land_use_change_co2_per_capita: tonnes per person (t/person); methane: million tonnes CO2-equivalent (GWP100); methane_per_capita: t CO2-equivalent per person; nitrous_oxide: million tonnes CO2-equivalent (GWP100); nitrous_oxide_per_capita: t CO2-equivalent per person; oil_co2: million tonnes (Mt); oil_co2_per_capita: tonnes per person (t/person); other_co2_per_capita: tonnes per person (t/person); other_industry_co2: million tonnes (Mt); primary_energy_consumption: terawatt-hours (TWh); share_global_cement_co2: %; share_global_co2: %; share_global_co2_including_luc: %; share_global_coal_co2: %; share_global_cumulative_cement_co2: %; share_global_cumulative_co2: %; share_global_cumulative_co2_including_luc: %; share_global_cumulative_coal_co2: %; share_global_cumulative_flaring_co2: %; share_global_cumulative_gas_co2: %; share_global_cumulative_luc_co2: %; share_global_cumulative_oil_co2: %; share_global_cumulative_other_co2: %; share_global_flaring_co2: %; share_global_gas_co2: %; share_global_luc_co2: %; share_global_oil_co2: %; share_global_other_co2: %; share_of_temperature_change_from_ghg: %; temperature_change_from_ch4: °C; temperature_change_from_co2: °C; temperature_change_from_ghg: °C; temperature_change_from_n2o: °C; total_ghg: million tonnes CO2-equivalent (GWP100); total_ghg_excluding_lucf: million tonnes CO2-equivalent (GWP100); trade_co2: million tonnes CO2 (Mt); trade_co2_share: %
- Supplied
- 29 Oct 2026, 14:35
- Checksum
- sha256 7ec10de1e9502c20…
Result (RESULT_JSON)
{
"sample": {
"iso_entities": 218,
"countries": 119,
"country_years": 3905,
"kuwait_rows_dropped": 3,
"terciles_n": {
"low": 40,
"high": 40,
"mid": 39
},
"tercile_cutoffs_gdppc_1990": [
4695,
12349.3
],
"accounting_max_abs_gap_mt": 0.00048828125
},
"long_differences": {
"1990-2007": {
"n": 116,
"co2": {
"low": {
"b": 0.8787,
"se": 0.2954,
"p": 0.0029,
"stars": "***"
},
"mid": {
"b": 0.7483,
"se": 0.1261,
"p": 0,
"stars": "***"
},
"high": {
"b": 0.7455,
"se": 0.1785,
"p": 0,
"stars": "***"
},
"r2": 0.5124
},
"consumption_co2": {
"low": {
"b": 0.6163,
"se": 0.2418,
"p": 0.0108,
"stars": "**"
},
"mid": {
"b": 0.4145,
"se": 0.1032,
"p": 0.0001,
"stars": "***"
},
"high": {
"b": 0.7729,
"se": 0.1117,
"p": 0,
"stars": "***"
},
"r2": 0.3844
},
"wedge": {
"low": {
"b": -0.2624,
"se": 0.2494,
"p": 0.2927,
"stars": ""
},
"mid": {
"b": -0.3338,
"se": 0.1351,
"p": 0.0135,
"stars": "**"
},
"high": {
"b": 0.0274,
"se": 0.1174,
"p": 0.8155,
"stars": ""
},
"r2": 0.1192
},
"wedge_high_minus_low": {
"b": 0.2898,
"se": 0.2757,
"p": 0.2931
},
"mean_dlog": {
"high": {
"gdp": 0.6403,
"co2": 0.1661,
"consumption_co2": 0.287,
"wedge": 0.121
},
"low": {
"gdp": 0.7526,
"co2": 0.7024,
"consumption_co2": 0.8601,
"wedge": 0.1577
},
"mid": {
"gdp": 0.7345,
"co2": 0.327,
"consumption_co2": 0.2893,
"wedge": -0.0377
}
},
"mean_wedge_test": {
"low": {
"mean": 0.1577,
"median": 0.0593,
"se": 0.0659,
"p": 0.0166
},
"mid": {
"mean": -0.0377,
"median": 0.0072,
"se": 0.1024,
"p": 0.7129
},
"high": {
"mean": 0.121,
"median": 0.0612,
"se": 0.0392,
"p": 0.0021
}
},
"high_dlog_co2_minus_cons": -0.121,
"high_wedge_wls_baseline_co2": {
"b": 0.0665,
"se": 0.1031,
"p": 0.519,
"stars": ""
},
"high_abs_decoupled_territorial": 11,
"high_abs_decoupled_consumption": 8,
"high_n": 38
},
"2007-2022": {
"n": 119,
"co2": {
"low": {
"b": 1.1307,
"se": 0.3669,
"p": 0.0021,
"stars": "***"
},
"mid": {
"b": 1.0013,
"se": 0.1396,
"p": 0,
"stars": "***"
},
"high": {
"b": 0.4521,
"se": 0.2278,
"p": 0.0472,
"stars": "**"
},
"r2": 0.6813
},
"consumption_co2": {
"low": {
"b": 1.1165,
"se": 0.2291,
"p": 0,
"stars": "***"
},
"mid": {
"b": 1.3201,
"se": 0.6451,
"p": 0.0407,
"stars": "**"
},
"high": {
"b": 0.6387,
"se": 0.3173,
"p": 0.0441,
"stars": "**"
},
"r2": 0.4509
},
"wedge": {
"low": {
"b": -0.0142,
"se": 0.372,
"p": 0.9695,
"stars": ""
},
"mid": {
"b": 0.3187,
"se": 0.635,
"p": 0.6157,
"stars": ""
},
"high": {
"b": 0.1866,
"se": 0.2128,
"p": 0.3805,
"stars": ""
},
"r2": 0.0378
},
"wedge_high_minus_low": {
"b": 0.2008,
"se": 0.4286,
"p": 0.6393
},
"mean_dlog": {
"high": {
"gdp": 0.2248,
"co2": -0.2239,
"consumption_co2": -0.1199,
"wedge": 0.104
},
"low": {
"gdp": 0.717,
"co2": 0.8258,
"consumption_co2": 0.8054,
"wedge": -0.0204
},
"mid": {
"gdp": 0.4105,
"co2": 0.1874,
"consumption_co2": 0.3597,
"wedge": 0.1723
}
},
"mean_wedge_test": {
"low": {
"mean": -0.0204,
"median": -0.0301,
"se": 0.0757,
"p": 0.7875
},
"mid": {
"mean": 0.1723,
"median": 0.0484,
"se": 0.1033,
"p": 0.0953
},
"high": {
"mean": 0.104,
"median": 0.0756,
"se": 0.0481,
"p": 0.0308
}
},
"high_dlog_co2_minus_cons": -0.104,
"high_wedge_wls_baseline_co2": {
"b": -0.0432,
"se": 0.0332,
"p": 0.1926,
"stars": ""
},
"high_abs_decoupled_territorial": 32,
"high_abs_decoupled_consumption": 27,
"high_n": 40
},
"stacked": {
"change_high_wedge": 0.1592,
"se": 0.2475,
"p": 0.5201,
"n_obs": 235,
"n_countries": 119
}
},
"annual_first_differences": {
"n_obs": 3782,
"n_countries": 119,
"co2": {
"r2": 0.1901,
"low_pre": {
"dk": {
"b": 0.6974,
"se": 0.0822,
"p": 0,
"stars": "***"
},
"cl_se": 0.1681
},
"low_post": {
"dk": {
"b": 1.2959,
"se": 0.1154,
"p": 0,
"stars": "***"
},
"cl_se": 0.0871
},
"mid_pre": {
"dk": {
"b": 0.5972,
"se": 0.0932,
"p": 0,
"stars": "***"
},
"cl_se": 0.075
},
"mid_post": {
"dk": {
"b": 0.904,
"se": 0.144,
"p": 0,
"stars": "***"
},
"cl_se": 0.1219
},
"high_pre": {
"dk": {
"b": 0.5112,
"se": 0.1737,
"p": 0.0032,
"stars": "***"
},
"cl_se": 0.1103
},
"high_post": {
"dk": {
"b": 0.5313,
"se": 0.0821,
"p": 0,
"stars": "***"
},
"cl_se": 0.0604
}
},
"consumption_co2": {
"r2": 0.0777,
"low_pre": {
"dk": {
"b": 0.6138,
"se": 0.0827,
"p": 0,
"stars": "***"
},
"cl_se": 0.1762
},
"low_post": {
"dk": {
"b": 1.2161,
"se": 0.2063,
"p": 0,
"stars": "***"
},
"cl_se": 0.1022
},
"mid_pre": {
"dk": {
"b": 0.5087,
"se": 0.0997,
"p": 0,
"stars": "***"
},
"cl_se": 0.0881
},
"mid_post": {
"dk": {
"b": 0.7653,
"se": 0.232,
"p": 0.001,
"stars": "***"
},
"cl_se": 0.1167
},
"high_pre": {
"dk": {
"b": 0.4977,
"se": 0.1142,
"p": 0,
"stars": "***"
},
"cl_se": 0.0833
},
"high_post": {
"dk": {
"b": 0.5173,
"se": 0.3957,
"p": 0.1911,
"stars": ""
},
"cl_se": 0.2079
}
},
"wedge": {
"r2": 0.015,
"low_pre": {
"dk": {
"b": -0.0836,
"se": 0.0566,
"p": 0.1399,
"stars": ""
},
"cl_se": 0.0616
},
"low_post": {
"dk": {
"b": -0.0798,
"se": 0.164,
"p": 0.6265,
"stars": ""
},
"cl_se": 0.103
},
"mid_pre": {
"dk": {
"b": -0.0884,
"se": 0.0942,
"p": 0.3476,
"stars": ""
},
"cl_se": 0.0639
},
"mid_post": {
"dk": {
"b": -0.1387,
"se": 0.3012,
"p": 0.6452,
"stars": ""
},
"cl_se": 0.1415
},
"high_pre": {
"dk": {
"b": -0.0135,
"se": 0.1363,
"p": 0.921,
"stars": ""
},
"cl_se": 0.0996
},
"high_post": {
"dk": {
"b": -0.014,
"se": 0.4185,
"p": 0.9733,
"stars": ""
},
"cl_se": 0.2114
}
},
"wedge_high_change_dk": {
"b": -0.0005,
"se": 0.44,
"p": 0.9991
}
},
"trade_shares": {
"1990": {
"mean_trade_co2_share": 19.231,
"median_trade_co2_share": 14.045,
"aggregate_trade_over_co2_pct": 7.386,
"aggregate_trade_co2_mt": 863.5,
"n": 38,
"share_net_importers": 0.737
},
"2007": {
"mean_trade_co2_share": 36.769,
"median_trade_co2_share": 24.297,
"aggregate_trade_over_co2_pct": 14.428,
"aggregate_trade_co2_mt": 2019.4,
"n": 40,
"share_net_importers": 0.825
},
"2022": {
"mean_trade_co2_share": 70.542,
"median_trade_co2_share": 24.373,
"aggregate_trade_over_co2_pct": 18.163,
"aggregate_trade_co2_mt": 2149,
"n": 40,
"share_net_importers": 0.825
},
"2022_low_mean_trade_co2_share": 34.868
},
"robustness": {
"start1995_high_wedge_pre": {
"b": 0.1009,
"se": 0.0957,
"p": 0.2915,
"stars": "",
"n": 118
},
"pop1m_high_wedge_pre": {
"b": 0.0343,
"se": 0.1175,
"p": 0.7705,
"stars": "",
"n": 111
},
"percapita_high_wedge_pre": {
"b": 0.1456,
"se": 0.2095,
"p": 0.487,
"stars": ""
},
"pop1m_high_wedge_post": {
"b": -0.0087,
"se": 0.0535,
"p": 0.8711,
"stars": "",
"n": 114
},
"percapita_high_wedge_post": {
"b": 0.208,
"se": 0.2087,
"p": 0.3188,
"stars": ""
},
"fe_levels_high_pre": {
"b": 0.0874,
"se": 0.0832,
"p": 0.2934,
"stars": ""
},
"fe_levels_high_post": {
"b": 0.0569,
"se": 0.0778,
"p": 0.4642,
"stars": ""
},
"fe_levels_n": 3905
},
"revision": {
"kuwait_facts": {
"co2_1990": 37.86,
"co2_1991": 492.98,
"co2_1992": 29.68,
"gdp_change_1990_91_pct": -71.4,
"oil_fires_entity_nonmissing_co2": 0,
"tercile": "high"
},
"exploratory_replication": {
"rich_x_dgdp": {
"b": 1.3181,
"se": 0.5787,
"p": 0.0227,
"stars": "**"
},
"dgdp_rest": {
"b": -0.0281,
"se": 0.06,
"p": 0.6392,
"stars": ""
},
"rich_total": 1.29,
"n_obs": 3739,
"rich_x_dgdp_without_kuwait": {
"b": 0.0799,
"se": 0.1291,
"p": 0.5359,
"stars": ""
}
},
"kuwait_variants": {
"drop1990_92": {
"n_countries": 119,
"n_obs": 3905,
"ld_n": [
116,
119
],
"ld_high_co2": {
"pre": {
"b": 0.7455,
"se": 0.1785,
"p": 0,
"stars": "***"
},
"post": {
"b": 0.4521,
"se": 0.2278,
"p": 0.0472,
"stars": "**"
}
},
"ld_high_consumption_co2": {
"pre": {
"b": 0.7729,
"se": 0.1117,
"p": 0,
"stars": "***"
},
"post": {
"b": 0.6387,
"se": 0.3173,
"p": 0.0441,
"stars": "**"
}
},
"ld_high_wedge": {
"pre": {
"b": 0.0274,
"se": 0.1174,
"p": 0.8155,
"stars": ""
},
"post": {
"b": 0.1866,
"se": 0.2128,
"p": 0.3805,
"stars": ""
}
},
"ld_stacked_change": {
"change_high_wedge": 0.1592,
"se": 0.2475,
"p": 0.5201,
"n_obs": 235,
"n_countries": 119
},
"ld_weighted": {
"pre": {
"b": 0.0665,
"se": 0.1031,
"p": 0.519,
"stars": ""
},
"post": {
"b": -0.0432,
"se": 0.0332,
"p": 0.1926,
"stars": ""
}
},
"fd_high_wedge": {
"pre": {
"b": -0.0135,
"se": 0.1363,
"p": 0.921,
"stars": "",
"cl_se": 0.0996
},
"post": {
"b": -0.014,
"se": 0.4185,
"p": 0.9733,
"stars": "",
"cl_se": 0.2114
}
},
"start1995_pre": {
"b": 0.1009,
"se": 0.0957,
"p": 0.2915,
"stars": ""
},
"pop1m": {
"pre": {
"b": 0.0343,
"se": 0.1175,
"p": 0.7705,
"stars": ""
},
"post": {
"b": -0.0087,
"se": 0.0535,
"p": 0.8711,
"stars": ""
}
},
"percapita": {
"pre": {
"b": 0.1456,
"se": 0.2095,
"p": 0.487,
"stars": ""
},
"post": {
"b": 0.208,
"se": 0.2087,
"p": 0.3188,
"stars": ""
}
},
"fe_levels": {
"pre": {
"b": 0.0874,
"se": 0.0832,
"p": 0.2934,
"stars": ""
},
"post": {
"b": 0.0569,
"se": 0.0778,
"p": 0.4642,
"stars": ""
}
}
},
"keep_all": {
"n_countries": 119,
"n_obs": 3908,
"ld_n": [
117,
119
],
"ld_high_co2": {
"pre": {
"b": 0.6372,
"se": 0.1325,
"p": 0,
"stars": "***"
},
"post": {
"b": 0.4521,
"se": 0.2278,
"p": 0.0472,
"stars": "**"
}
},
"ld_high_consumption_co2": {
"pre": {
"b": 0.6392,
"se": 0.1139,
"p": 0,
"stars": "***"
},
"post": {
"b": 0.6387,
"se": 0.3173,
"p": 0.0441,
"stars": "**"
}
},
"ld_high_wedge": {
"pre": {
"b": 0.002,
"se": 0.0884,
"p": 0.9817,
"stars": ""
},
"post": {
"b": 0.1866,
"se": 0.2128,
"p": 0.3805,
"stars": ""
}
},
"ld_stacked_change": {
"change_high_wedge": 0.1846,
"se": 0.2373,
"p": 0.4366,
"n_obs": 236,
"n_countries": 119
},
"ld_weighted": {
"pre": {
"b": 0.0522,
"se": 0.0884,
"p": 0.555,
"stars": ""
},
"post": {
"b": -0.0432,
"se": 0.0332,
"p": 0.1926,
"stars": ""
}
},
"fd_high_wedge": {
"pre": {
"b": 1.4753,
"se": 0.4282,
"p": 0.0006,
"stars": "***",
"cl_se": 0.551
},
"post": {
"b": -0.014,
"se": 0.4185,
"p": 0.9733,
"stars": "",
"cl_se": 0.2114
}
},
"start1995_pre": {
"b": 0.1009,
"se": 0.0957,
"p": 0.2915,
"stars": ""
},
"pop1m": {
"pre": {
"b": 0.0061,
"se": 0.0885,
"p": 0.945,
"stars": ""
},
"post": {
"b": -0.0087,
"se": 0.0535,
"p": 0.8711,
"stars": ""
}
},
"percapita": {
"pre": {
"b": 0.0403,
"se": 0.1248,
"p": 0.7469,
"stars": ""
},
"post": {
"b": 0.208,
"se": 0.2087,
"p": 0.3188,
"stars": ""
}
},
"fe_levels": {
"pre": {
"b": 0.1713,
"se": 0.0982,
"p": 0.0811,
"stars": "*"
},
"post": {
"b": 0.1397,
"se": 0.0944,
"p": 0.1388,
"stars": ""
}
}
},
"drop1991_92": {
"n_countries": 119,
"n_obs": 3906,
"ld_n": [
117,
119
],
"ld_high_co2": {
"pre": {
"b": 0.6372,
"se": 0.1325,
"p": 0,
"stars": "***"
},
"post": {
"b": 0.4521,
"se": 0.2278,
"p": 0.0472,
"stars": "**"
}
},
"ld_high_consumption_co2": {
"pre": {
"b": 0.6392,
"se": 0.1139,
"p": 0,
"stars": "***"
},
"post": {
"b": 0.6387,
"se": 0.3173,
"p": 0.0441,
"stars": "**"
}
},
"ld_high_wedge": {
"pre": {
"b": 0.002,
"se": 0.0884,
"p": 0.9817,
"stars": ""
},
"post": {
"b": 0.1866,
"se": 0.2128,
"p": 0.3805,
"stars": ""
}
},
"ld_stacked_change": {
"change_high_wedge": 0.1846,
"se": 0.2373,
"p": 0.4366,
"n_obs": 236,
"n_countries": 119
},
"ld_weighted": {
"pre": {
"b": 0.0522,
"se": 0.0884,
"p": 0.555,
"stars": ""
},
"post": {
"b": -0.0432,
"se": 0.0332,
"p": 0.1926,
"stars": ""
}
},
"fd_high_wedge": {
"pre": {
"b": -0.0135,
"se": 0.1363,
"p": 0.921,
"stars": "",
"cl_se": 0.0996
},
"post": {
"b": -0.014,
"se": 0.4185,
"p": 0.9733,
"stars": "",
"cl_se": 0.2114
}
},
"start1995_pre": {
"b": 0.1009,
"se": 0.0957,
"p": 0.2915,
"stars": ""
},
"pop1m": {
"pre": {
"b": 0.0061,
"se": 0.0885,
"p": 0.945,
"stars": ""
},
"post": {
"b": -0.0087,
"se": 0.0535,
"p": 0.8711,
"stars": ""
}
},
"percapita": {
"pre": {
"b": 0.0403,
"se": 0.1248,
"p": 0.7469,
"stars": ""
},
"post": {
"b": 0.208,
"se": 0.2087,
"p": 0.3188,
"stars": ""
}
},
"fe_levels": {
"pre": {
"b": 0.0819,
"se": 0.0808,
"p": 0.3106,
"stars": ""
},
"post": {
"b": 0.0515,
"se": 0.0755,
"p": 0.495,
"stars": ""
}
}
}
},
"leave_one_out": {
"ld_pre": {
"min": -0.0533,
"argmin": "Singapore",
"max": 0.0893,
"argmax": "Latvia"
},
"ld_post": {
"min": -0.0075,
"argmin": "Malta",
"max": 0.3753,
"argmax": "Venezuela"
},
"wls_pre": {
"min": -0.0063,
"argmin": "Japan",
"max": 0.138,
"argmax": "Saudi Arabia"
},
"wls_post": {
"min": -0.0615,
"argmin": "Venezuela",
"max": -0.0222,
"argmax": "United Arab Emirates"
},
"fd_pre": {
"min": -0.0516,
"argmin": "Singapore",
"max": 0.0556,
"argmax": "Lithuania"
},
"fd_post": {
"min": -0.0784,
"argmin": "Malta",
"max": 0.2596,
"argmax": "Venezuela"
},
"n_fits": 119
},
"intercepts": {
"1990-2007": {
"low": {
"intercept_demeaned": {
"b": 0.1577,
"se": 0.0649,
"p": 0.0151,
"stars": "**"
},
"intercept_raw": {
"b": 0.3552,
"se": 0.2073,
"p": 0.0866,
"stars": "*"
},
"mean_dlog_gdp": 0.7526,
"growth_component": -0.1975,
"growth_share_of_mean": -1.2519
},
"mid": {
"intercept_demeaned": {
"b": -0.0377,
"se": 0.0976,
"p": 0.6994,
"stars": ""
},
"intercept_raw": {
"b": 0.2075,
"se": 0.1159,
"p": 0.0733,
"stars": "*"
},
"mean_dlog_gdp": 0.7345,
"growth_component": -0.2452,
"growth_share_of_mean": 6.5043
},
"high": {
"intercept_demeaned": {
"b": 0.121,
"se": 0.0397,
"p": 0.0023,
"stars": "***"
},
"intercept_raw": {
"b": 0.1034,
"se": 0.0827,
"p": 0.2113,
"stars": ""
},
"mean_dlog_gdp": 0.6403,
"growth_component": 0.0175,
"growth_share_of_mean": 0.145
},
"high_weighted": {
"intercept_demeaned": {
"b": 0.0726,
"se": 0.0163,
"p": 0,
"stars": "***"
},
"intercept_raw": {
"b": 0.039,
"se": 0.0565,
"p": 0.4903,
"stars": ""
},
"slope": {
"b": 0.0665,
"se": 0.1031,
"p": 0.519,
"stars": ""
},
"weighted_mean_dlog_gdp": 0.5058
}
},
"2007-2022": {
"low": {
"intercept_demeaned": {
"b": -0.0204,
"se": 0.0767,
"p": 0.7902,
"stars": ""
},
"intercept_raw": {
"b": -0.0102,
"se": 0.2407,
"p": 0.9662,
"stars": ""
},
"mean_dlog_gdp": 0.717,
"growth_component": -0.0102,
"growth_share_of_mean": 0.5001
},
"mid": {
"intercept_demeaned": {
"b": 0.1723,
"se": 0.1039,
"p": 0.0972,
"stars": "*"
},
"intercept_raw": {
"b": 0.0415,
"se": 0.1882,
"p": 0.8257,
"stars": ""
},
"mean_dlog_gdp": 0.4105,
"growth_component": 0.1308,
"growth_share_of_mean": 0.7594
},
"high": {
"intercept_demeaned": {
"b": 0.104,
"se": 0.048,
"p": 0.0302,
"stars": "**"
},
"intercept_raw": {
"b": 0.0621,
"se": 0.0399,
"p": 0.1197,
"stars": ""
},
"mean_dlog_gdp": 0.2248,
"growth_component": 0.0419,
"growth_share_of_mean": 0.4033
},
"high_weighted": {
"intercept_demeaned": {
"b": 0.0365,
"se": 0.0098,
"p": 0.0002,
"stars": "***"
},
"intercept_raw": {
"b": 0.0458,
"se": 0.012,
"p": 0.0001,
"stars": "***"
},
"slope": {
"b": -0.0432,
"se": 0.0332,
"p": 0.1926,
"stars": ""
},
"weighted_mean_dlog_gdp": 0.2141
}
}
},
"aggregate_wedge_high": {
"balanced_1990_2007": {
"n": 38,
"w_start": 0.0713,
"w_end": 0.1351,
"change": 0.0638
},
"balanced_2007_2022": {
"n": 40,
"w_start": 0.1348,
"w_end": 0.1669,
"change": 0.0321
},
"table1_implied_1990_2007": 0.0635,
"table1_implied_2007_2022": 0.0321
},
"endpoints": {
"avg3_1990-92_to_2005-07": {
"n": 114,
"high_n": 37,
"co2": {
"b": 0.8657,
"se": 0.2929,
"p": 0.0031,
"stars": "***"
},
"consumption_co2": {
"b": 0.9121,
"se": 0.2009,
"p": 0,
"stars": "***"
},
"wedge": {
"b": 0.0464,
"se": 0.2491,
"p": 0.8522,
"stars": ""
},
"wedge_weighted": {
"b": 0.093,
"se": 0.1584,
"p": 0.557,
"stars": ""
},
"high_mean_wedge": 0.0817,
"high_median_wedge": 0.0203
},
"avg3_2005-07_to_2020-22": {
"n": 118,
"high_n": 40,
"co2": {
"b": 0.5508,
"se": 0.2133,
"p": 0.0098,
"stars": "***"
},
"consumption_co2": {
"b": 0.647,
"se": 0.2975,
"p": 0.0297,
"stars": "**"
},
"wedge": {
"b": 0.0962,
"se": 0.2093,
"p": 0.6458,
"stars": ""
},
"wedge_weighted": {
"b": -0.0756,
"se": 0.0572,
"p": 0.1862,
"stars": ""
},
"high_mean_wedge": 0.1016,
"high_median_wedge": 0.0798
},
"end2019_2007-2019": {
"n": 119,
"high_n": 40,
"co2": {
"b": 0.5552,
"se": 0.2174,
"p": 0.0106,
"stars": "**"
},
"consumption_co2": {
"b": 1.0908,
"se": 0.1953,
"p": 0,
"stars": "***"
},
"wedge": {
"b": 0.5355,
"se": 0.2745,
"p": 0.0511,
"stars": "*"
},
"wedge_weighted": {
"b": 0.238,
"se": 0.1747,
"p": 0.1729,
"stars": ""
},
"high_mean_wedge": 0.0373,
"high_median_wedge": -0.001
},
"benchmark_1997-2007": {
"n": 118,
"high_n": 39,
"co2": {
"b": 0.4974,
"se": 0.1308,
"p": 0.0001,
"stars": "***"
},
"consumption_co2": {
"b": 0.5547,
"se": 0.1161,
"p": 0,
"stars": "***"
},
"wedge": {
"b": 0.0574,
"se": 0.12,
"p": 0.6325,
"stars": ""
},
"wedge_weighted": {
"b": 0.1369,
"se": 0.1387,
"p": 0.3235,
"stars": ""
},
"high_mean_wedge": 0.1208,
"high_median_wedge": 0.0932
},
"benchmark_2007-2017": {
"n": 119,
"high_n": 40,
"co2": {
"b": 0.6128,
"se": 0.2116,
"p": 0.0038,
"stars": "***"
},
"consumption_co2": {
"b": 1.0498,
"se": 0.2293,
"p": 0,
"stars": "***"
},
"wedge": {
"b": 0.437,
"se": 0.3443,
"p": 0.2042,
"stars": ""
},
"wedge_weighted": {
"b": 0.0804,
"se": 0.0902,
"p": 0.373,
"stars": ""
},
"high_mean_wedge": 0.0233,
"high_median_wedge": -0.0052
}
},
"consumption_status_gcb2025": {
"1990": "extended_back",
"1992": "extended_back",
"1995": "extended_back",
"1997": "benchmark",
"2005": "between_benchmarks",
"2006": "between_benchmarks",
"2007": "benchmark",
"2017": "benchmark",
"2019": "extended_forward",
"2020": "extended_forward",
"2022": "extended_forward"
},
"ci95": {
"ld_high_wedge_pre": [
-0.2027,
0.2575
],
"ld_high_wedge_post": [
-0.2305,
0.6037
],
"stacked_change": [
-0.3259,
0.6443
],
"weighted_pre": [
-0.1356,
0.2686
],
"weighted_post": [
-0.1083,
0.0219
],
"fe_levels_pre": [
-0.0757,
0.2505
],
"fe_levels_post": [
-0.0956,
0.2094
]
},
"round2_drift": {
"1997-2007": {
"n_high": 39,
"mean_wedge": 0.1208,
"median_wedge": 0.0932,
"mean_dlog_gdp": 0.4495,
"mean_dlog_co2": 0.1356,
"mean_dlog_cons": 0.2563,
"intercept_demeaned": {
"b": 0.1208,
"se": 0.0338,
"p": 0.0004,
"stars": "***"
},
"intercept_raw": {
"b": 0.095,
"se": 0.0633,
"p": 0.1336,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.064,
"se": 0.0152,
"p": 0,
"stars": "***"
},
"weighted_intercept_raw": {
"b": 0.0209,
"se": 0.0506,
"p": 0.6794,
"stars": ""
},
"aggregate_balanced": {
"n": 39,
"w_start": 0.0738,
"w_end": 0.1342,
"change": 0.0604
}
},
"2007-2017": {
"n_high": 40,
"mean_wedge": 0.0233,
"median_wedge": -0.0052,
"mean_dlog_gdp": 0.159,
"mean_dlog_co2": -0.1099,
"mean_dlog_cons": -0.0866,
"intercept_demeaned": {
"b": 0.0233,
"se": 0.0433,
"p": 0.591,
"stars": ""
},
"intercept_raw": {
"b": -0.0462,
"se": 0.0382,
"p": 0.2263,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": -0.0181,
"se": 0.0117,
"p": 0.1206,
"stars": ""
},
"weighted_intercept_raw": {
"b": -0.03,
"se": 0.0165,
"p": 0.0697,
"stars": "*"
},
"aggregate_balanced": {
"n": 40,
"w_start": 0.1348,
"w_end": 0.1139,
"change": -0.0209
}
},
"2007-2019": {
"n_high": 40,
"mean_wedge": 0.0373,
"median_wedge": -0.001,
"mean_dlog_gdp": 0.1916,
"mean_dlog_co2": -0.1502,
"mean_dlog_cons": -0.1129,
"intercept_demeaned": {
"b": 0.0373,
"se": 0.0455,
"p": 0.412,
"stars": ""
},
"intercept_raw": {
"b": -0.0653,
"se": 0.0371,
"p": 0.0781,
"stars": "*"
},
"weighted_intercept_demeaned": {
"b": -0.0107,
"se": 0.0112,
"p": 0.339,
"stars": ""
},
"weighted_intercept_raw": {
"b": -0.0539,
"se": 0.0365,
"p": 0.1401,
"stars": ""
},
"aggregate_balanced": {
"n": 40,
"w_start": 0.1348,
"w_end": 0.1231,
"change": -0.0117
}
},
"2017-2022": {
"n_high": 40,
"mean_wedge": 0.0807,
"median_wedge": 0.0884,
"mean_dlog_gdp": 0.0658,
"mean_dlog_co2": -0.1141,
"mean_dlog_cons": -0.0334,
"intercept_demeaned": {
"b": 0.0807,
"se": 0.0199,
"p": 0.0001,
"stars": "***"
},
"intercept_raw": {
"b": 0.087,
"se": 0.0231,
"p": 0.0002,
"stars": "***"
},
"weighted_intercept_demeaned": {
"b": 0.0524,
"se": 0.0089,
"p": 0,
"stars": "***"
},
"weighted_intercept_raw": {
"b": 0.0658,
"se": 0.0083,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 40,
"w_start": 0.1139,
"w_end": 0.1669,
"change": 0.053
}
},
"2007-2022_check": {
"n_high": 40,
"mean_wedge": 0.104,
"median_wedge": 0.0756,
"mean_dlog_gdp": 0.2248,
"mean_dlog_co2": -0.2239,
"mean_dlog_cons": -0.1199,
"intercept_demeaned": {
"b": 0.104,
"se": 0.048,
"p": 0.0302,
"stars": "**"
},
"intercept_raw": {
"b": 0.0621,
"se": 0.0399,
"p": 0.1197,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0365,
"se": 0.0098,
"p": 0.0002,
"stars": "***"
},
"weighted_intercept_raw": {
"b": 0.0458,
"se": 0.012,
"p": 0.0001,
"stars": "***"
},
"aggregate_balanced": {
"n": 40,
"w_start": 0.1348,
"w_end": 0.1669,
"change": 0.0321
}
}
},
"round3": {
"members": {
"low": [
"Albania",
"Bangladesh",
"Benin",
"Bolivia",
"Burkina Faso",
"Cambodia",
"Cameroon",
"China",
"Cote d'Ivoire",
"Dominican Republic",
"Egypt",
"El Salvador",
"Ethiopia",
"Ghana",
"Guinea",
"Honduras",
"India",
"Indonesia",
"Kenya",
"Laos",
"Madagascar",
"Malawi",
"Mongolia",
"Morocco",
"Mozambique",
"Nepal",
"Nicaragua",
"Nigeria",
"Pakistan",
"Philippines",
"Rwanda",
"Senegal",
"Sri Lanka",
"Tajikistan",
"Tanzania",
"Togo",
"Uganda",
"Vietnam",
"Zambia",
"Zimbabwe"
],
"mid": [
"Argentina",
"Armenia",
"Azerbaijan",
"Bahrain",
"Belarus",
"Botswana",
"Brazil",
"Bulgaria",
"Chile",
"Colombia",
"Costa Rica",
"Croatia",
"Ecuador",
"Georgia",
"Guatemala",
"Hungary",
"Iran",
"Jamaica",
"Jordan",
"Kazakhstan",
"Kyrgyzstan",
"Malaysia",
"Mauritius",
"Mexico",
"Namibia",
"Oman",
"Panama",
"Paraguay",
"Peru",
"Poland",
"Qatar",
"Romania",
"Russia",
"South Africa",
"Thailand",
"Tunisia",
"Turkey",
"Ukraine",
"Uruguay"
],
"high": [
"Australia",
"Austria",
"Belgium",
"Canada",
"Cyprus",
"Czechia",
"Denmark",
"Estonia",
"Finland",
"France",
"Germany",
"Greece",
"Hong Kong",
"Ireland",
"Israel",
"Italy",
"Japan",
"Kuwait",
"Latvia",
"Lithuania",
"Luxembourg",
"Malta",
"Netherlands",
"New Zealand",
"Norway",
"Portugal",
"Saudi Arabia",
"Singapore",
"Slovakia",
"Slovenia",
"South Korea",
"Spain",
"Sweden",
"Switzerland",
"Taiwan",
"Trinidad and Tobago",
"United Arab Emirates",
"United Kingdom",
"United States",
"Venezuela"
]
},
"variant_members": {
"all": [
"Australia",
"Austria",
"Belgium",
"Canada",
"Cyprus",
"Czechia",
"Denmark",
"Estonia",
"Finland",
"France",
"Germany",
"Greece",
"Hong Kong",
"Ireland",
"Israel",
"Italy",
"Japan",
"Kuwait",
"Latvia",
"Lithuania",
"Luxembourg",
"Malta",
"Netherlands",
"New Zealand",
"Norway",
"Portugal",
"Saudi Arabia",
"Singapore",
"Slovakia",
"Slovenia",
"South Korea",
"Spain",
"Sweden",
"Switzerland",
"Taiwan",
"Trinidad and Tobago",
"United Arab Emirates",
"United Kingdom",
"United States",
"Venezuela"
],
"no_oil_transition": [
"Australia",
"Austria",
"Belgium",
"Canada",
"Cyprus",
"Denmark",
"Finland",
"France",
"Germany",
"Greece",
"Hong Kong",
"Ireland",
"Israel",
"Italy",
"Japan",
"Luxembourg",
"Malta",
"Netherlands",
"New Zealand",
"Norway",
"Portugal",
"Singapore",
"South Korea",
"Spain",
"Sweden",
"Switzerland",
"Taiwan",
"United Kingdom",
"United States"
],
"net_importers_1997": [
"Austria",
"Canada",
"Denmark",
"Finland",
"France",
"Germany",
"Greece",
"Hong Kong",
"Ireland",
"Israel",
"Italy",
"Japan",
"Latvia",
"Lithuania",
"Malta",
"Netherlands",
"New Zealand",
"Portugal",
"Singapore",
"Slovenia",
"South Korea",
"Spain",
"Sweden",
"Switzerland",
"Taiwan",
"United Arab Emirates",
"United Kingdom"
]
},
"high_net_exporters_1997": [
"Australia",
"Belgium",
"Cyprus",
"Czechia",
"Estonia",
"Kuwait",
"Luxembourg",
"Saudi Arabia",
"Slovakia",
"Trinidad and Tobago",
"United States",
"Venezuela"
],
"high_net_exporters_2007": [
"Australia",
"Czechia",
"Kuwait",
"Luxembourg",
"Saudi Arabia",
"Trinidad and Tobago",
"Venezuela"
],
"groups": {
"all": {
"1990-2007": {
"n_high": 38,
"n_total": 116,
"slope_co2_hc1": {
"b": 0.7455,
"se": 0.1785,
"p": 0,
"stars": "***",
"ci95": [
0.3956,
1.0954
]
},
"slope_consumption_co2_hc1": {
"b": 0.7729,
"se": 0.1117,
"p": 0,
"stars": "***",
"ci95": [
0.554,
0.9918
]
},
"slope_wedge_hc1": {
"b": 0.0274,
"se": 0.1174,
"p": 0.8155,
"stars": "",
"ci95": [
-0.2027,
0.2575
]
},
"slope_wedge_hc3": {
"b": 0.0274,
"se_hc3": 0.1375,
"p_hc3": 0.8421,
"ci95_hc3": [
-0.2421,
0.2969
]
},
"slope_wedge_wild_p": 0.8086,
"slope_wedge_weighted_hc1": {
"b": 0.0665,
"se": 0.1031,
"p": 0.519,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": 0.0665,
"se_hc3": 0.1254,
"p_hc3": 0.5959,
"ci95_hc3": [
-0.1793,
0.3123
]
},
"slope_wedge_weighted_wild_p": 0.5634,
"kish_n_weighted": 4.55,
"mean_dlog_gdp": 0.6403,
"mean_wedge": 0.121,
"median_wedge": 0.0612,
"intercept_demeaned": {
"b": 0.121,
"se": 0.0397,
"p": 0.0023,
"stars": "***"
},
"intercept_raw": {
"b": 0.1034,
"se": 0.0827,
"p": 0.2113,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0726,
"se": 0.0163,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 38,
"w_start": 0.0713,
"w_end": 0.1351,
"change": 0.0638
},
"upper_ci_times_mean_growth": 0.1649,
"growth_share_point": 0.145,
"growth_share_ci": [
-1.0726,
1.3626
],
"mde_80pct_5pct": 0.3289
},
"2007-2022": {
"n_high": 40,
"n_total": 119,
"slope_co2_hc1": {
"b": 0.4521,
"se": 0.2278,
"p": 0.0472,
"stars": "**",
"ci95": [
0.0056,
0.8986
]
},
"slope_consumption_co2_hc1": {
"b": 0.6387,
"se": 0.3173,
"p": 0.0441,
"stars": "**",
"ci95": [
0.0168,
1.2606
]
},
"slope_wedge_hc1": {
"b": 0.1866,
"se": 0.2128,
"p": 0.3805,
"stars": "",
"ci95": [
-0.2305,
0.6037
]
},
"slope_wedge_hc3": {
"b": 0.1866,
"se_hc3": 0.2798,
"p_hc3": 0.5048,
"ci95_hc3": [
-0.3618,
0.735
]
},
"slope_wedge_wild_p": 0.5005,
"slope_wedge_weighted_hc1": {
"b": -0.0432,
"se": 0.0332,
"p": 0.1926,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": -0.0432,
"se_hc3": 0.0385,
"p_hc3": 0.2615,
"ci95_hc3": [
-0.1187,
0.0322
]
},
"slope_wedge_weighted_wild_p": 0.1495,
"kish_n_weighted": 4.67,
"mean_dlog_gdp": 0.2248,
"mean_wedge": 0.104,
"median_wedge": 0.0756,
"intercept_demeaned": {
"b": 0.104,
"se": 0.048,
"p": 0.0302,
"stars": "**"
},
"intercept_raw": {
"b": 0.0621,
"se": 0.0399,
"p": 0.1197,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0365,
"se": 0.0098,
"p": 0.0002,
"stars": "***"
},
"aggregate_balanced": {
"n": 40,
"w_start": 0.1348,
"w_end": 0.1669,
"change": 0.0321
},
"upper_ci_times_mean_growth": 0.1357,
"growth_share_point": 0.4033,
"growth_share_ci": [
-0.4982,
1.3049
],
"mde_80pct_5pct": 0.5962
},
"1997-2007": {
"n_high": 39,
"n_total": 118,
"slope_co2_hc1": {
"b": 0.4974,
"se": 0.1308,
"p": 0.0001,
"stars": "***",
"ci95": [
0.241,
0.7538
]
},
"slope_consumption_co2_hc1": {
"b": 0.5547,
"se": 0.1161,
"p": 0,
"stars": "***",
"ci95": [
0.3271,
0.7823
]
},
"slope_wedge_hc1": {
"b": 0.0574,
"se": 0.12,
"p": 0.6325,
"stars": "",
"ci95": [
-0.1778,
0.2926
]
},
"slope_wedge_hc3": {
"b": 0.0574,
"se_hc3": 0.1256,
"p_hc3": 0.6479,
"ci95_hc3": [
-0.1889,
0.3036
]
},
"slope_wedge_wild_p": 0.6125,
"slope_wedge_weighted_hc1": {
"b": 0.1369,
"se": 0.1387,
"p": 0.3235,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": 0.1369,
"se_hc3": 0.182,
"p_hc3": 0.4519,
"ci95_hc3": [
-0.2199,
0.4937
]
},
"slope_wedge_weighted_wild_p": 0.5036,
"kish_n_weighted": 4.47,
"mean_dlog_gdp": 0.4495,
"mean_wedge": 0.1208,
"median_wedge": 0.0932,
"intercept_demeaned": {
"b": 0.1208,
"se": 0.0338,
"p": 0.0004,
"stars": "***"
},
"intercept_raw": {
"b": 0.095,
"se": 0.0633,
"p": 0.1336,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.064,
"se": 0.0152,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 39,
"w_start": 0.0738,
"w_end": 0.1342,
"change": 0.0604
},
"upper_ci_times_mean_growth": 0.1315,
"growth_share_point": 0.2136,
"growth_share_ci": [
-0.6616,
1.0888
],
"mde_80pct_5pct": 0.3362
},
"2007-2017": {
"n_high": 40,
"n_total": 119,
"slope_co2_hc1": {
"b": 0.6128,
"se": 0.2116,
"p": 0.0038,
"stars": "***",
"ci95": [
0.1981,
1.0275
]
},
"slope_consumption_co2_hc1": {
"b": 1.0498,
"se": 0.2293,
"p": 0,
"stars": "***",
"ci95": [
0.6004,
1.4992
]
},
"slope_wedge_hc1": {
"b": 0.437,
"se": 0.3443,
"p": 0.2042,
"stars": "",
"ci95": [
-0.2378,
1.1118
]
},
"slope_wedge_hc3": {
"b": 0.437,
"se_hc3": 0.3788,
"p_hc3": 0.2486,
"ci95_hc3": [
-0.3055,
1.1796
]
},
"slope_wedge_wild_p": 0.2127,
"slope_wedge_weighted_hc1": {
"b": 0.0804,
"se": 0.0902,
"p": 0.373,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": 0.0804,
"se_hc3": 0.1104,
"p_hc3": 0.4665,
"ci95_hc3": [
-0.1359,
0.2966
]
},
"slope_wedge_weighted_wild_p": 0.4562,
"kish_n_weighted": 4.67,
"mean_dlog_gdp": 0.159,
"mean_wedge": 0.0233,
"median_wedge": -0.0052,
"intercept_demeaned": {
"b": 0.0233,
"se": 0.0433,
"p": 0.591,
"stars": ""
},
"intercept_raw": {
"b": -0.0462,
"se": 0.0382,
"p": 0.2263,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": -0.0181,
"se": 0.0117,
"p": 0.1206,
"stars": ""
},
"aggregate_balanced": {
"n": 40,
"w_start": 0.1348,
"w_end": 0.1139,
"change": -0.0209
},
"upper_ci_times_mean_growth": 0.1768,
"growth_share_point": 2.9821,
"growth_share_ci": [
-1.6228,
7.587
],
"mde_80pct_5pct": 0.9646
},
"2017-2022": {
"n_high": 40,
"n_total": 119,
"slope_co2_hc1": {
"b": 0.2137,
"se": 0.0782,
"p": 0.0063,
"stars": "***",
"ci95": [
0.0604,
0.367
]
},
"slope_consumption_co2_hc1": {
"b": 0.1175,
"se": 0.1186,
"p": 0.3221,
"stars": "",
"ci95": [
-0.115,
0.35
]
},
"slope_wedge_hc1": {
"b": -0.0962,
"se": 0.0887,
"p": 0.2784,
"stars": "",
"ci95": [
-0.27,
0.0776
]
},
"slope_wedge_hc3": {
"b": -0.0962,
"se_hc3": 0.2811,
"p_hc3": 0.7322,
"ci95_hc3": [
-0.6471,
0.4547
]
},
"slope_wedge_wild_p": 0.5581,
"slope_wedge_weighted_hc1": {
"b": -0.1945,
"se": 0.0396,
"p": 0,
"stars": "***"
},
"slope_wedge_weighted_hc3": {
"b": -0.1945,
"se_hc3": 0.0885,
"p_hc3": 0.0279,
"ci95_hc3": [
-0.3679,
-0.0211
]
},
"slope_wedge_weighted_wild_p": 0.1892,
"kish_n_weighted": 5.17,
"mean_dlog_gdp": 0.0658,
"mean_wedge": 0.0807,
"median_wedge": 0.0884,
"intercept_demeaned": {
"b": 0.0807,
"se": 0.0199,
"p": 0.0001,
"stars": "***"
},
"intercept_raw": {
"b": 0.087,
"se": 0.0231,
"p": 0.0002,
"stars": "***"
},
"weighted_intercept_demeaned": {
"b": 0.0524,
"se": 0.0089,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 40,
"w_start": 0.1139,
"w_end": 0.1669,
"change": 0.053
},
"upper_ci_times_mean_growth": 0.0051,
"growth_share_point": -0.0784,
"growth_share_ci": [
-0.2201,
0.0633
],
"mde_80pct_5pct": 0.2485
}
},
"no_oil_transition": {
"1990-2007": {
"n_high": 28,
"n_total": 106,
"slope_co2_hc1": {
"b": 0.474,
"se": 0.2752,
"p": 0.085,
"stars": "*",
"ci95": [
-0.0654,
1.0134
]
},
"slope_consumption_co2_hc1": {
"b": 0.7795,
"se": 0.1991,
"p": 0.0001,
"stars": "***",
"ci95": [
0.3893,
1.1697
]
},
"slope_wedge_hc1": {
"b": 0.3055,
"se": 0.2337,
"p": 0.1912,
"stars": "",
"ci95": [
-0.1525,
0.7635
]
},
"slope_wedge_hc3": {
"b": 0.3055,
"se_hc3": 0.2946,
"p_hc3": 0.2998,
"ci95_hc3": [
-0.272,
0.883
]
},
"slope_wedge_wild_p": 0.3367,
"slope_wedge_weighted_hc1": {
"b": 0.2034,
"se": 0.168,
"p": 0.226,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": 0.2034,
"se_hc3": 0.2097,
"p_hc3": 0.3323,
"ci95_hc3": [
-0.2077,
0.6144
]
},
"slope_wedge_weighted_wild_p": 0.3613,
"kish_n_weighted": 4.03,
"mean_dlog_gdp": 0.6136,
"mean_wedge": 0.1207,
"median_wedge": 0.0398,
"intercept_demeaned": {
"b": 0.1207,
"se": 0.0429,
"p": 0.0049,
"stars": "***"
},
"intercept_raw": {
"b": -0.0667,
"se": 0.1421,
"p": 0.6387,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0726,
"se": 0.0158,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 28,
"w_start": 0.0792,
"w_end": 0.1458,
"change": 0.0666
},
"upper_ci_times_mean_growth": 0.4685,
"growth_share_point": 1.5531,
"growth_share_ci": [
-0.7753,
3.8814
],
"mde_80pct_5pct": 0.6547
},
"2007-2022": {
"n_high": 29,
"n_total": 108,
"slope_co2_hc1": {
"b": 0.7551,
"se": 0.2438,
"p": 0.002,
"stars": "***",
"ci95": [
0.2773,
1.2329
]
},
"slope_consumption_co2_hc1": {
"b": 1.4933,
"se": 0.3415,
"p": 0,
"stars": "***",
"ci95": [
0.824,
2.1626
]
},
"slope_wedge_hc1": {
"b": 0.7383,
"se": 0.5506,
"p": 0.18,
"stars": "",
"ci95": [
-0.3409,
1.8175
]
},
"slope_wedge_hc3": {
"b": 0.7383,
"se_hc3": 0.6775,
"p_hc3": 0.2758,
"ci95_hc3": [
-0.5895,
2.0661
]
},
"slope_wedge_wild_p": 0.2132,
"slope_wedge_weighted_hc1": {
"b": -0.0522,
"se": 0.0826,
"p": 0.5273,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": -0.0522,
"se_hc3": 0.0945,
"p_hc3": 0.5807,
"ci95_hc3": [
-0.2375,
0.1331
]
},
"slope_wedge_weighted_wild_p": 0.5297,
"kish_n_weighted": 4.01,
"mean_dlog_gdp": 0.2496,
"mean_wedge": 0.1215,
"median_wedge": 0.0778,
"intercept_demeaned": {
"b": 0.1215,
"se": 0.0612,
"p": 0.0471,
"stars": "**"
},
"intercept_raw": {
"b": -0.0628,
"se": 0.1111,
"p": 0.5722,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0362,
"se": 0.0103,
"p": 0.0004,
"stars": "***"
},
"aggregate_balanced": {
"n": 29,
"w_start": 0.1464,
"w_end": 0.1845,
"change": 0.0381
},
"upper_ci_times_mean_growth": 0.4536,
"growth_share_point": 1.5167,
"growth_share_ci": [
-0.7003,
3.7337
],
"mde_80pct_5pct": 1.5426
},
"1997-2007": {
"n_high": 28,
"n_total": 107,
"slope_co2_hc1": {
"b": 0.3025,
"se": 0.2844,
"p": 0.2875,
"stars": "",
"ci95": [
-0.2549,
0.8599
]
},
"slope_consumption_co2_hc1": {
"b": 0.6275,
"se": 0.1998,
"p": 0.0017,
"stars": "***",
"ci95": [
0.2359,
1.0191
]
},
"slope_wedge_hc1": {
"b": 0.325,
"se": 0.3423,
"p": 0.3424,
"stars": "",
"ci95": [
-0.3459,
0.9959
]
},
"slope_wedge_hc3": {
"b": 0.325,
"se_hc3": 0.378,
"p_hc3": 0.3898,
"ci95_hc3": [
-0.4158,
1.0658
]
},
"slope_wedge_wild_p": 0.3646,
"slope_wedge_weighted_hc1": {
"b": 0.3771,
"se": 0.2248,
"p": 0.0934,
"stars": "*"
},
"slope_wedge_weighted_hc3": {
"b": 0.3771,
"se_hc3": 0.403,
"p_hc3": 0.3495,
"ci95_hc3": [
-0.4129,
1.167
]
},
"slope_wedge_weighted_wild_p": 0.468,
"kish_n_weighted": 3.98,
"mean_dlog_gdp": 0.3637,
"mean_wedge": 0.1325,
"median_wedge": 0.0925,
"intercept_demeaned": {
"b": 0.1325,
"se": 0.0407,
"p": 0.0011,
"stars": "***"
},
"intercept_raw": {
"b": 0.0143,
"se": 0.1285,
"p": 0.9116,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0653,
"se": 0.0132,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 28,
"w_start": 0.0818,
"w_end": 0.1458,
"change": 0.064
},
"upper_ci_times_mean_growth": 0.3622,
"growth_share_point": 0.8921,
"growth_share_ci": [
-0.9495,
2.7337
],
"mde_80pct_5pct": 0.959
},
"2007-2017": {
"n_high": 29,
"n_total": 108,
"slope_co2_hc1": {
"b": 0.5617,
"se": 0.3191,
"p": 0.0784,
"stars": "*",
"ci95": [
-0.0637,
1.1871
]
},
"slope_consumption_co2_hc1": {
"b": 1.5469,
"se": 0.3739,
"p": 0,
"stars": "***",
"ci95": [
0.8141,
2.2797
]
},
"slope_wedge_hc1": {
"b": 0.9852,
"se": 0.6631,
"p": 0.1373,
"stars": "",
"ci95": [
-0.3145,
2.2849
]
},
"slope_wedge_hc3": {
"b": 0.9852,
"se_hc3": 0.7999,
"p_hc3": 0.2181,
"ci95_hc3": [
-0.5826,
2.5531
]
},
"slope_wedge_wild_p": 0.0675,
"slope_wedge_weighted_hc1": {
"b": 0.0519,
"se": 0.1628,
"p": 0.7499,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": 0.0519,
"se_hc3": 0.1873,
"p_hc3": 0.7817,
"ci95_hc3": [
-0.3152,
0.4189
]
},
"slope_wedge_weighted_wild_p": 0.7802,
"kish_n_weighted": 4.01,
"mean_dlog_gdp": 0.1596,
"mean_wedge": 0.034,
"median_wedge": -0.0067,
"intercept_demeaned": {
"b": 0.034,
"se": 0.0539,
"p": 0.5282,
"stars": ""
},
"intercept_raw": {
"b": -0.1232,
"se": 0.0809,
"p": 0.1277,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": -0.0202,
"se": 0.0126,
"p": 0.1092,
"stars": ""
},
"aggregate_balanced": {
"n": 29,
"w_start": 0.1464,
"w_end": 0.1273,
"change": -0.0191
},
"upper_ci_times_mean_growth": 0.3647,
"growth_share_point": 4.6246,
"growth_share_ci": [
-1.4763,
10.7256
],
"mde_80pct_5pct": 1.8577
},
"2017-2022": {
"n_high": 29,
"n_total": 108,
"slope_co2_hc1": {
"b": 1.126,
"se": 0.2489,
"p": 0,
"stars": "***",
"ci95": [
0.6382,
1.6138
]
},
"slope_consumption_co2_hc1": {
"b": 0.5652,
"se": 0.4221,
"p": 0.1805,
"stars": "",
"ci95": [
-0.2621,
1.3925
]
},
"slope_wedge_hc1": {
"b": -0.5608,
"se": 0.4211,
"p": 0.183,
"stars": "",
"ci95": [
-1.3861,
0.2645
]
},
"slope_wedge_hc3": {
"b": -0.5608,
"se_hc3": 0.4925,
"p_hc3": 0.2549,
"ci95_hc3": [
-1.5261,
0.4046
]
},
"slope_wedge_wild_p": 0.2547,
"slope_wedge_weighted_hc1": {
"b": -0.342,
"se": 0.1697,
"p": 0.0439,
"stars": "**"
},
"slope_wedge_weighted_hc3": {
"b": -0.342,
"se_hc3": 0.2671,
"p_hc3": 0.2004,
"ci95_hc3": [
-0.8655,
0.1815
]
},
"slope_wedge_weighted_wild_p": 0.1395,
"kish_n_weighted": 4.21,
"mean_dlog_gdp": 0.09,
"mean_wedge": 0.0875,
"median_wedge": 0.0887,
"intercept_demeaned": {
"b": 0.0875,
"se": 0.02,
"p": 0,
"stars": "***"
},
"intercept_raw": {
"b": 0.138,
"se": 0.0515,
"p": 0.0074,
"stars": "***"
},
"weighted_intercept_demeaned": {
"b": 0.0564,
"se": 0.0095,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 29,
"w_start": 0.1273,
"w_end": 0.1845,
"change": 0.0572
},
"upper_ci_times_mean_growth": 0.0238,
"growth_share_point": -0.5768,
"growth_share_ci": [
-1.4257,
0.2721
],
"mde_80pct_5pct": 1.1798
}
},
"net_importers_1997": {
"1990-2007": {
"n_high": 27,
"n_total": 105,
"slope_co2_hc1": {
"b": 0.6268,
"se": 0.1659,
"p": 0.0002,
"stars": "***",
"ci95": [
0.3016,
0.952
]
},
"slope_consumption_co2_hc1": {
"b": 0.7571,
"se": 0.1268,
"p": 0,
"stars": "***",
"ci95": [
0.5086,
1.0056
]
},
"slope_wedge_hc1": {
"b": 0.1303,
"se": 0.1185,
"p": 0.2715,
"stars": "",
"ci95": [
-0.102,
0.3626
]
},
"slope_wedge_hc3": {
"b": 0.1303,
"se_hc3": 0.1377,
"p_hc3": 0.344,
"ci95_hc3": [
-0.1396,
0.4002
]
},
"slope_wedge_wild_p": 0.1831,
"slope_wedge_weighted_hc1": {
"b": 0.1716,
"se": 0.1052,
"p": 0.1028,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": 0.1716,
"se_hc3": 0.1244,
"p_hc3": 0.1676,
"ci95_hc3": [
-0.0721,
0.4153
]
},
"slope_wedge_weighted_wild_p": 0.1759,
"kish_n_weighted": 8.7,
"mean_dlog_gdp": 0.6292,
"mean_wedge": 0.142,
"median_wedge": 0.0727,
"intercept_demeaned": {
"b": 0.142,
"se": 0.0466,
"p": 0.0023,
"stars": "***"
},
"intercept_raw": {
"b": 0.06,
"se": 0.09,
"p": 0.5049,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0594,
"se": 0.0282,
"p": 0.0351,
"stars": "**"
},
"aggregate_balanced": {
"n": 27,
"w_start": 0.1707,
"w_end": 0.2246,
"change": 0.0539
},
"upper_ci_times_mean_growth": 0.2281,
"growth_share_point": 0.5774,
"growth_share_ci": [
-0.452,
1.6067
],
"mde_80pct_5pct": 0.332
},
"2007-2022": {
"n_high": 27,
"n_total": 106,
"slope_co2_hc1": {
"b": 0.8562,
"se": 0.2317,
"p": 0.0002,
"stars": "***",
"ci95": [
0.4021,
1.3103
]
},
"slope_consumption_co2_hc1": {
"b": 1.3832,
"se": 0.2936,
"p": 0,
"stars": "***",
"ci95": [
0.8078,
1.9586
]
},
"slope_wedge_hc1": {
"b": 0.527,
"se": 0.4994,
"p": 0.2913,
"stars": "",
"ci95": [
-0.4518,
1.5058
]
},
"slope_wedge_hc3": {
"b": 0.527,
"se_hc3": 0.5981,
"p_hc3": 0.3783,
"ci95_hc3": [
-0.6453,
1.6993
]
},
"slope_wedge_wild_p": 0.4849,
"slope_wedge_weighted_hc1": {
"b": -0.0894,
"se": 0.0972,
"p": 0.3574,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": -0.0894,
"se_hc3": 0.1204,
"p_hc3": 0.4578,
"ci95_hc3": [
-0.3254,
0.1466
]
},
"slope_wedge_weighted_wild_p": 0.4546,
"kish_n_weighted": 10.14,
"mean_dlog_gdp": 0.2563,
"mean_wedge": 0.1367,
"median_wedge": 0.0785,
"intercept_demeaned": {
"b": 0.1367,
"se": 0.0647,
"p": 0.0346,
"stars": "**"
},
"intercept_raw": {
"b": 0.0016,
"se": 0.0928,
"p": 0.9865,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0385,
"se": 0.0203,
"p": 0.0581,
"stars": "*"
},
"aggregate_balanced": {
"n": 27,
"w_start": 0.2246,
"w_end": 0.2686,
"change": 0.044
},
"upper_ci_times_mean_growth": 0.3859,
"growth_share_point": 0.9881,
"growth_share_ci": [
-0.8471,
2.8232
],
"mde_80pct_5pct": 1.3991
},
"1997-2007": {
"n_high": 27,
"n_total": 106,
"slope_co2_hc1": {
"b": 0.3662,
"se": 0.1676,
"p": 0.0289,
"stars": "**",
"ci95": [
0.0377,
0.6947
]
},
"slope_consumption_co2_hc1": {
"b": 0.6448,
"se": 0.0894,
"p": 0,
"stars": "***",
"ci95": [
0.4696,
0.82
]
},
"slope_wedge_hc1": {
"b": 0.2786,
"se": 0.1497,
"p": 0.0628,
"stars": "*",
"ci95": [
-0.0148,
0.572
]
},
"slope_wedge_hc3": {
"b": 0.2786,
"se_hc3": 0.2603,
"p_hc3": 0.2845,
"ci95_hc3": [
-0.2316,
0.7888
]
},
"slope_wedge_wild_p": 0.0908,
"slope_wedge_weighted_hc1": {
"b": 0.3078,
"se": 0.1414,
"p": 0.0295,
"stars": "**"
},
"slope_wedge_weighted_hc3": {
"b": 0.3078,
"se_hc3": 0.2194,
"p_hc3": 0.1606,
"ci95_hc3": [
-0.1222,
0.7377
]
},
"slope_wedge_weighted_wild_p": 0.2224,
"kish_n_weighted": 9.48,
"mean_dlog_gdp": 0.4146,
"mean_wedge": 0.1298,
"median_wedge": 0.1004,
"intercept_demeaned": {
"b": 0.1298,
"se": 0.0373,
"p": 0.0005,
"stars": "***"
},
"intercept_raw": {
"b": 0.0143,
"se": 0.0679,
"p": 0.8329,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0499,
"se": 0.0231,
"p": 0.0307,
"stars": "**"
},
"aggregate_balanced": {
"n": 27,
"w_start": 0.173,
"w_end": 0.2246,
"change": 0.0516
},
"upper_ci_times_mean_growth": 0.2372,
"growth_share_point": 0.8899,
"growth_share_ci": [
-0.0473,
1.8271
],
"mde_80pct_5pct": 0.4194
},
"2007-2017": {
"n_high": 27,
"n_total": 106,
"slope_co2_hc1": {
"b": 0.6644,
"se": 0.2484,
"p": 0.0075,
"stars": "***",
"ci95": [
0.1775,
1.1513
]
},
"slope_consumption_co2_hc1": {
"b": 1.29,
"se": 0.3075,
"p": 0,
"stars": "***",
"ci95": [
0.6873,
1.8927
]
},
"slope_wedge_hc1": {
"b": 0.6256,
"se": 0.5391,
"p": 0.2459,
"stars": "",
"ci95": [
-0.431,
1.6822
]
},
"slope_wedge_hc3": {
"b": 0.6256,
"se_hc3": 0.6564,
"p_hc3": 0.3406,
"ci95_hc3": [
-0.661,
1.9121
]
},
"slope_wedge_wild_p": 0.3462,
"slope_wedge_weighted_hc1": {
"b": -0.0648,
"se": 0.113,
"p": 0.5661,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": -0.0648,
"se_hc3": 0.1371,
"p_hc3": 0.6363,
"ci95_hc3": [
-0.3336,
0.2039
]
},
"slope_wedge_weighted_wild_p": 0.5773,
"kish_n_weighted": 10.14,
"mean_dlog_gdp": 0.1646,
"mean_wedge": 0.0303,
"median_wedge": -0.0262,
"intercept_demeaned": {
"b": 0.0303,
"se": 0.0608,
"p": 0.6177,
"stars": ""
},
"intercept_raw": {
"b": -0.0726,
"se": 0.0567,
"p": 0.2002,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": -0.0429,
"se": 0.0157,
"p": 0.0062,
"stars": "***"
},
"aggregate_balanced": {
"n": 27,
"w_start": 0.2246,
"w_end": 0.1848,
"change": -0.0398
},
"upper_ci_times_mean_growth": 0.2769,
"growth_share_point": 3.3985,
"growth_share_ci": [
-2.3413,
9.1383
],
"mde_80pct_5pct": 1.5103
},
"2017-2022": {
"n_high": 27,
"n_total": 106,
"slope_co2_hc1": {
"b": 1.0824,
"se": 0.2578,
"p": 0,
"stars": "***",
"ci95": [
0.5771,
1.5877
]
},
"slope_consumption_co2_hc1": {
"b": 0.6561,
"se": 0.3651,
"p": 0.0724,
"stars": "*",
"ci95": [
-0.0595,
1.3717
]
},
"slope_wedge_hc1": {
"b": -0.4263,
"se": 0.4184,
"p": 0.3082,
"stars": "",
"ci95": [
-1.2463,
0.3937
]
},
"slope_wedge_hc3": {
"b": -0.4263,
"se_hc3": 0.4903,
"p_hc3": 0.3846,
"ci95_hc3": [
-1.3874,
0.5347
]
},
"slope_wedge_wild_p": 0.5308,
"slope_wedge_weighted_hc1": {
"b": -0.0468,
"se": 0.1622,
"p": 0.773,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": -0.0468,
"se_hc3": 0.2014,
"p_hc3": 0.8163,
"ci95_hc3": [
-0.4415,
0.3479
]
},
"slope_wedge_weighted_wild_p": 0.7751,
"kish_n_weighted": 9.9,
"mean_dlog_gdp": 0.0917,
"mean_wedge": 0.1063,
"median_wedge": 0.0891,
"intercept_demeaned": {
"b": 0.1063,
"se": 0.0168,
"p": 0,
"stars": "***"
},
"intercept_raw": {
"b": 0.1454,
"se": 0.0506,
"p": 0.0041,
"stars": "***"
},
"weighted_intercept_demeaned": {
"b": 0.0802,
"se": 0.0121,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 27,
"w_start": 0.1848,
"w_end": 0.2686,
"change": 0.0838
},
"upper_ci_times_mean_growth": 0.0361,
"growth_share_point": -0.3677,
"growth_share_ci": [
-1.0751,
0.3396
],
"mde_80pct_5pct": 1.1722
}
}
},
"weighted_vs_unweighted": {
"1990-2007": {
"diff_weighted_minus_unweighted": 0.0391,
"boot_se": 0.1243,
"boot_ci95_pct": [
-0.1455,
0.3704
],
"dumouchel_duncan_gdp_x_w": {
"b": 0.0408,
"se": 0.3502,
"p": 0.9072,
"stars": "",
"se_type": "HC3"
},
"n": 38
},
"2007-2022": {
"diff_weighted_minus_unweighted": -0.2298,
"boot_se": 0.2819,
"boot_ci95_pct": [
-1.0327,
0.0246
],
"dumouchel_duncan_gdp_x_w": {
"b": -0.4445,
"se": 0.7494,
"p": 0.5531,
"stars": "",
"se_type": "HC3"
},
"n": 40
}
},
"fd_group_period_intercepts": {
"n_obs": 3782,
"co2": {
"r2": 0.2006,
"low_pre": {
"dk": {
"b": 0.5324,
"se": 0.1022,
"p": 0,
"stars": "***"
},
"cl_se": 0.2075
},
"low_post": {
"dk": {
"b": 0.9871,
"se": 0.156,
"p": 0,
"stars": "***"
},
"cl_se": 0.1368
},
"mid_pre": {
"dk": {
"b": 0.6274,
"se": 0.0993,
"p": 0,
"stars": "***"
},
"cl_se": 0.082
},
"mid_post": {
"dk": {
"b": 0.8594,
"se": 0.1431,
"p": 0,
"stars": "***"
},
"cl_se": 0.1417
},
"high_pre": {
"dk": {
"b": 0.6614,
"se": 0.1791,
"p": 0.0002,
"stars": "***"
},
"cl_se": 0.132
},
"high_post": {
"dk": {
"b": 0.5768,
"se": 0.0655,
"p": 0,
"stars": "***"
},
"cl_se": 0.0676
}
},
"consumption_co2": {
"r2": 0.0833,
"low_pre": {
"dk": {
"b": 0.383,
"se": 0.0882,
"p": 0,
"stars": "***"
},
"cl_se": 0.2031
},
"low_post": {
"dk": {
"b": 1.0861,
"se": 0.2314,
"p": 0,
"stars": "***"
},
"cl_se": 0.1456
},
"mid_pre": {
"dk": {
"b": 0.5728,
"se": 0.0814,
"p": 0,
"stars": "***"
},
"cl_se": 0.0957
},
"mid_post": {
"dk": {
"b": 0.6833,
"se": 0.2025,
"p": 0.0007,
"stars": "***"
},
"cl_se": 0.1207
},
"high_pre": {
"dk": {
"b": 0.654,
"se": 0.1026,
"p": 0,
"stars": "***"
},
"cl_se": 0.0903
},
"high_post": {
"dk": {
"b": 0.5667,
"se": 0.4272,
"p": 0.1846,
"stars": ""
},
"cl_se": 0.2463
}
},
"wedge": {
"r2": 0.016,
"low_pre": {
"dk": {
"b": -0.1494,
"se": 0.1056,
"p": 0.1573,
"stars": ""
},
"cl_se": 0.0779
},
"low_post": {
"dk": {
"b": 0.099,
"se": 0.1324,
"p": 0.4545,
"stars": ""
},
"cl_se": 0.1425
},
"mid_pre": {
"dk": {
"b": -0.0546,
"se": 0.1237,
"p": 0.6589,
"stars": ""
},
"cl_se": 0.071
},
"mid_post": {
"dk": {
"b": -0.1761,
"se": 0.2691,
"p": 0.5129,
"stars": ""
},
"cl_se": 0.1688
},
"high_pre": {
"dk": {
"b": -0.0074,
"se": 0.2078,
"p": 0.9715,
"stars": ""
},
"cl_se": 0.1338
},
"high_post": {
"dk": {
"b": -0.0101,
"se": 0.43,
"p": 0.9813,
"stars": ""
},
"cl_se": 0.2191
}
},
"wedge_high_intercepts_dk": {
"g_high_0": {
"b": -0.0083,
"se": 0.0136,
"p": 0.5412,
"stars": ""
},
"g_high_1": {
"b": 0.0132,
"se": 0.0093,
"p": 0.1561,
"stars": ""
}
}
},
"levels_fe": {
"n_obs": 3905,
"group_period_fe": {
"pre": {
"b": 0.0815,
"se": 0.079,
"p": 0.3022,
"stars": ""
},
"post": {
"b": 0.0426,
"se": 0.0697,
"p": 0.5407,
"stars": ""
}
},
"country_trends": {
"pre": {
"b": 0.0327,
"se": 0.1058,
"p": 0.7572,
"stars": ""
},
"post": {
"b": 0.0446,
"se": 0.1058,
"p": 0.6734,
"stars": ""
}
}
},
"benchmark_signature": {
"kink_median_abs_second_diff": {
"bench_mean": 0.051,
"other_mean": 0.0599,
"perm_p_greater": 0.8307,
"bench_ranks_of_n": {
"1997": 21,
"2001": 14,
"2004": 23,
"2007": 15,
"2011": 8,
"2014": 16,
"2017": 24
},
"n_years": 30
},
"dispersion_sd_dw": {
"bench_mean": 0.1413,
"other_mean": 0.1441,
"perm_p_greater": 0.5401,
"bench_ranks_of_n": {
"1997": 18,
"2001": 9,
"2004": 25,
"2007": 13,
"2011": 3,
"2014": 16,
"2017": 27
},
"n_years": 30
},
"top5_kink_years": [
1999,
2000,
2008,
2009,
1998
],
"top5_sd_years": [
2000,
2018,
2011,
2003,
1996
]
},
"china_trade_co2": {
"y1990": -122.8,
"y2005": -1235.4,
"y2007": -1506,
"y2022": -1284.8,
"years_negative": 33,
"years_observed": 33
}
},
"round4": {
"high_group_co2_shares": {
"1990": {
"n_members": 38,
"total_mt": 11691.7,
"oil_transition_pct": 6,
"oil_pct": 3.2,
"transition_pct": 2.8,
"united_states_pct": 43.9,
"australia_pct": 2.4,
"dropped_by_net_importers_1997_pct": 52.4
},
"1997": {
"n_members": 39,
"total_mt": 12710.2,
"oil_transition_pct": 5.8,
"oil_pct": 4,
"transition_pct": 1.8,
"united_states_pct": 44.6,
"australia_pct": 2.5,
"dropped_by_net_importers_1997_pct": 53.2
},
"2007": {
"n_members": 40,
"total_mt": 13996.7,
"oil_transition_pct": 7.6,
"oil_pct": 6,
"transition_pct": 1.6,
"united_states_pct": 43.7,
"australia_pct": 2.8,
"dropped_by_net_importers_1997_pct": 54.3
},
"2017": {
"n_members": 40,
"total_mt": 12664.3,
"oil_transition_pct": 10.4,
"oil_pct": 8.8,
"transition_pct": 1.6,
"united_states_pct": 41,
"australia_pct": 3.3,
"dropped_by_net_importers_1997_pct": 54.1
},
"2022": {
"n_members": 40,
"total_mt": 11831.4,
"oil_transition_pct": 11,
"oil_pct": 9.6,
"transition_pct": 1.4,
"united_states_pct": 42.7,
"australia_pct": 3.2,
"dropped_by_net_importers_1997_pct": 56.2
}
},
"note": "shares of the 1990 top tercile's summed territorial co2 in the analysis sample (Kuwait 1990-1992 rows excluded)"
},
"round5": {
"oil_rule_from_file": {
"rules_tried": {
"flaring_share_mean_1990_2022": {
"threshold_min_over_oil_set": 1.3592,
"selected": [
"Australia",
"Canada",
"Kuwait",
"New Zealand",
"Norway",
"Saudi Arabia",
"Trinidad and Tobago",
"United Arab Emirates",
"Venezuela"
],
"n_selected": 9,
"equals_oil_set": false,
"includes_norway": true,
"extra_beyond_oil_set": [
"Australia",
"Canada",
"New Zealand",
"Norway"
],
"norway_value": 3.9442,
"norway_rank_of_n": [
2,
40
]
},
"flaring_per_capita_mean_1990_2022": {
"threshold_min_over_oil_set": 0.2634,
"selected": [
"Australia",
"Canada",
"Kuwait",
"Norway",
"Saudi Arabia",
"Trinidad and Tobago",
"United Arab Emirates",
"Venezuela"
],
"n_selected": 8,
"equals_oil_set": false,
"includes_norway": true,
"extra_beyond_oil_set": [
"Australia",
"Canada",
"Norway"
],
"norway_value": 0.3541,
"norway_rank_of_n": [
6,
40
]
},
"flaring_share_1990": {
"threshold_min_over_oil_set": 0.2537,
"selected": [
"Australia",
"Canada",
"Czechia",
"Denmark",
"France",
"Germany",
"Italy",
"Kuwait",
"Netherlands",
"New Zealand",
"Norway",
"Saudi Arabia",
"Slovenia",
"Spain",
"Sweden",
"Trinidad and Tobago",
"United Arab Emirates",
"United Kingdom",
"United States",
"Venezuela"
],
"n_selected": 20,
"equals_oil_set": false,
"includes_norway": true,
"extra_beyond_oil_set": [
"Australia",
"Canada",
"Czechia",
"Denmark",
"France",
"Germany",
"Italy",
"Netherlands",
"New Zealand",
"Norway",
"Slovenia",
"Spain",
"Sweden",
"United Kingdom",
"United States"
],
"norway_value": 5.4738,
"norway_rank_of_n": [
3,
37
]
},
"flaring_share_1997": {
"threshold_min_over_oil_set": 1.6306,
"selected": [
"Australia",
"Canada",
"Kuwait",
"New Zealand",
"Norway",
"Saudi Arabia",
"Trinidad and Tobago",
"United Arab Emirates",
"Venezuela"
],
"n_selected": 9,
"equals_oil_set": false,
"includes_norway": true,
"extra_beyond_oil_set": [
"Australia",
"Canada",
"New Zealand",
"Norway"
],
"norway_value": 4.656,
"norway_rank_of_n": [
2,
40
]
},
"flaring_share_2007": {
"threshold_min_over_oil_set": 0.7627,
"selected": [
"Australia",
"Canada",
"Denmark",
"France",
"Kuwait",
"New Zealand",
"Norway",
"Portugal",
"Saudi Arabia",
"Slovenia",
"Sweden",
"Trinidad and Tobago",
"United Arab Emirates",
"United Kingdom",
"Venezuela"
],
"n_selected": 15,
"equals_oil_set": false,
"includes_norway": true,
"extra_beyond_oil_set": [
"Australia",
"Canada",
"Denmark",
"France",
"New Zealand",
"Norway",
"Portugal",
"Slovenia",
"Sweden",
"United Kingdom"
],
"norway_value": 6.177,
"norway_rank_of_n": [
1,
40
]
},
"net_exporter_embodied_1997": {
"selected": [
"Australia",
"Belgium",
"Cyprus",
"Czechia",
"Estonia",
"Kuwait",
"Luxembourg",
"Saudi Arabia",
"Slovakia",
"Trinidad and Tobago",
"United States",
"Venezuela"
],
"n_selected": 12,
"equals_oil_set": false,
"includes_norway": false,
"oil_set_missed": [
"United Arab Emirates"
],
"extra_beyond_oil_set": [
"Australia",
"Belgium",
"Cyprus",
"Czechia",
"Estonia",
"Luxembourg",
"Slovakia",
"United States"
]
},
"net_exporter_embodied_2007": {
"selected": [
"Australia",
"Czechia",
"Kuwait",
"Luxembourg",
"Saudi Arabia",
"Trinidad and Tobago",
"Venezuela"
],
"n_selected": 7,
"equals_oil_set": false,
"includes_norway": false,
"oil_set_missed": [
"United Arab Emirates"
],
"extra_beyond_oil_set": [
"Australia",
"Czechia",
"Luxembourg"
]
}
},
"any_rule_reproduces_oil_set": false,
"every_flaring_rule_includes_norway": true,
"norway_first_year_in_sample": 2003,
"note": "flaring rules: lowest threshold that keeps all five named oil exporters, over the 40 top-tercile countries, all file rows 1990-2022"
},
"core_no_norway": {
"1990-2007": {
"n_high": 28,
"n_total": 106,
"slope_co2_hc1": {
"b": 0.474,
"se": 0.2752,
"p": 0.085,
"stars": "*",
"ci95": [
-0.0654,
1.0134
]
},
"slope_consumption_co2_hc1": {
"b": 0.7795,
"se": 0.1991,
"p": 0.0001,
"stars": "***",
"ci95": [
0.3893,
1.1697
]
},
"slope_wedge_hc1": {
"b": 0.3055,
"se": 0.2337,
"p": 0.1912,
"stars": "",
"ci95": [
-0.1525,
0.7635
]
},
"slope_wedge_hc3": {
"b": 0.3055,
"se_hc3": 0.2946,
"p_hc3": 0.2998,
"ci95_hc3": [
-0.272,
0.883
]
},
"slope_wedge_wild_p": 0.3367,
"slope_wedge_weighted_hc1": {
"b": 0.2034,
"se": 0.168,
"p": 0.226,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": 0.2034,
"se_hc3": 0.2097,
"p_hc3": 0.3323,
"ci95_hc3": [
-0.2077,
0.6144
]
},
"slope_wedge_weighted_wild_p": 0.3613,
"kish_n_weighted": 4.03,
"mean_dlog_gdp": 0.6136,
"mean_wedge": 0.1207,
"median_wedge": 0.0398,
"intercept_demeaned": {
"b": 0.1207,
"se": 0.0429,
"p": 0.0049,
"stars": "***"
},
"intercept_raw": {
"b": -0.0667,
"se": 0.1421,
"p": 0.6387,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0726,
"se": 0.0158,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 28,
"w_start": 0.0792,
"w_end": 0.1458,
"change": 0.0666
},
"upper_ci_times_mean_growth": 0.4685,
"growth_share_point": 1.5531,
"growth_share_ci": [
-0.7753,
3.8814
],
"mde_80pct_5pct": 0.6547
},
"2007-2022": {
"n_high": 28,
"n_total": 107,
"slope_co2_hc1": {
"b": 0.7473,
"se": 0.2441,
"p": 0.0022,
"stars": "***",
"ci95": [
0.2689,
1.2257
]
},
"slope_consumption_co2_hc1": {
"b": 1.5283,
"se": 0.3321,
"p": 0,
"stars": "***",
"ci95": [
0.8774,
2.1792
]
},
"slope_wedge_hc1": {
"b": 0.7809,
"se": 0.5423,
"p": 0.1499,
"stars": "",
"ci95": [
-0.282,
1.8438
]
},
"slope_wedge_hc3": {
"b": 0.7809,
"se_hc3": 0.6698,
"p_hc3": 0.2437,
"ci95_hc3": [
-0.5319,
2.0938
]
},
"slope_wedge_wild_p": 0.1213,
"slope_wedge_weighted_hc1": {
"b": -0.0377,
"se": 0.082,
"p": 0.6454,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": -0.0377,
"se_hc3": 0.0943,
"p_hc3": 0.6891,
"ci95_hc3": [
-0.2227,
0.1472
]
},
"slope_wedge_weighted_wild_p": 0.6449,
"kish_n_weighted": 3.98,
"mean_dlog_gdp": 0.2473,
"mean_wedge": 0.1421,
"median_wedge": 0.0782,
"intercept_demeaned": {
"b": 0.1421,
"se": 0.0589,
"p": 0.0159,
"stars": "**"
},
"intercept_raw": {
"b": -0.0511,
"se": 0.1074,
"p": 0.6345,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0379,
"se": 0.0104,
"p": 0.0003,
"stars": "***"
},
"aggregate_balanced": {
"n": 28,
"w_start": 0.1458,
"w_end": 0.1856,
"change": 0.0398
},
"upper_ci_times_mean_growth": 0.456,
"growth_share_point": 1.359,
"growth_share_ci": [
-0.4908,
3.2088
],
"mde_80pct_5pct": 1.5193
},
"1997-2007": {
"n_high": 28,
"n_total": 107,
"slope_co2_hc1": {
"b": 0.3025,
"se": 0.2844,
"p": 0.2875,
"stars": "",
"ci95": [
-0.2549,
0.8599
]
},
"slope_consumption_co2_hc1": {
"b": 0.6275,
"se": 0.1998,
"p": 0.0017,
"stars": "***",
"ci95": [
0.2359,
1.0191
]
},
"slope_wedge_hc1": {
"b": 0.325,
"se": 0.3423,
"p": 0.3424,
"stars": "",
"ci95": [
-0.3459,
0.9959
]
},
"slope_wedge_hc3": {
"b": 0.325,
"se_hc3": 0.378,
"p_hc3": 0.3898,
"ci95_hc3": [
-0.4158,
1.0658
]
},
"slope_wedge_wild_p": 0.3646,
"slope_wedge_weighted_hc1": {
"b": 0.3771,
"se": 0.2248,
"p": 0.0934,
"stars": "*"
},
"slope_wedge_weighted_hc3": {
"b": 0.3771,
"se_hc3": 0.403,
"p_hc3": 0.3495,
"ci95_hc3": [
-0.4129,
1.167
]
},
"slope_wedge_weighted_wild_p": 0.468,
"kish_n_weighted": 3.98,
"mean_dlog_gdp": 0.3637,
"mean_wedge": 0.1325,
"median_wedge": 0.0925,
"intercept_demeaned": {
"b": 0.1325,
"se": 0.0407,
"p": 0.0011,
"stars": "***"
},
"intercept_raw": {
"b": 0.0143,
"se": 0.1285,
"p": 0.9116,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0653,
"se": 0.0132,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 28,
"w_start": 0.0818,
"w_end": 0.1458,
"change": 0.064
},
"upper_ci_times_mean_growth": 0.3622,
"growth_share_point": 0.8921,
"growth_share_ci": [
-0.9495,
2.7337
],
"mde_80pct_5pct": 0.959
},
"2007-2017": {
"n_high": 28,
"n_total": 107,
"slope_co2_hc1": {
"b": 0.5518,
"se": 0.3238,
"p": 0.0883,
"stars": "*",
"ci95": [
-0.0828,
1.1864
]
},
"slope_consumption_co2_hc1": {
"b": 1.5761,
"se": 0.3775,
"p": 0,
"stars": "***",
"ci95": [
0.8362,
2.316
]
},
"slope_wedge_hc1": {
"b": 1.0243,
"se": 0.6708,
"p": 0.1268,
"stars": "",
"ci95": [
-0.2904,
2.339
]
},
"slope_wedge_hc3": {
"b": 1.0243,
"se_hc3": 0.8158,
"p_hc3": 0.2092,
"ci95_hc3": [
-0.5746,
2.6231
]
},
"slope_wedge_wild_p": 0.0514,
"slope_wedge_weighted_hc1": {
"b": 0.0622,
"se": 0.165,
"p": 0.7062,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": 0.0622,
"se_hc3": 0.1902,
"p_hc3": 0.7436,
"ci95_hc3": [
-0.3105,
0.4349
]
},
"slope_wedge_weighted_wild_p": 0.7463,
"kish_n_weighted": 3.98,
"mean_dlog_gdp": 0.1568,
"mean_wedge": 0.0421,
"median_wedge": -0.0052,
"intercept_demeaned": {
"b": 0.0421,
"se": 0.0547,
"p": 0.4417,
"stars": ""
},
"intercept_raw": {
"b": -0.1186,
"se": 0.0803,
"p": 0.1396,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": -0.0196,
"se": 0.0125,
"p": 0.1164,
"stars": ""
},
"aggregate_balanced": {
"n": 28,
"w_start": 0.1458,
"w_end": 0.1274,
"change": -0.0185
},
"upper_ci_times_mean_growth": 0.3668,
"growth_share_point": 3.815,
"growth_share_ci": [
-1.0816,
8.7115
],
"mde_80pct_5pct": 1.8793
},
"2017-2022": {
"n_high": 28,
"n_total": 107,
"slope_co2_hc1": {
"b": 1.1336,
"se": 0.2514,
"p": 0,
"stars": "***",
"ci95": [
0.6409,
1.6263
]
},
"slope_consumption_co2_hc1": {
"b": 0.5247,
"se": 0.4019,
"p": 0.1917,
"stars": "",
"ci95": [
-0.263,
1.3124
]
},
"slope_wedge_hc1": {
"b": -0.609,
"se": 0.405,
"p": 0.1326,
"stars": "",
"ci95": [
-1.4028,
0.1848
]
},
"slope_wedge_hc3": {
"b": -0.609,
"se_hc3": 0.474,
"p_hc3": 0.1988,
"ci95_hc3": [
-1.538,
0.32
]
},
"slope_wedge_wild_p": 0.1571,
"slope_wedge_weighted_hc1": {
"b": -0.3423,
"se": 0.1745,
"p": 0.0498,
"stars": "**"
},
"slope_wedge_weighted_hc3": {
"b": -0.3423,
"se_hc3": 0.2838,
"p_hc3": 0.2278,
"ci95_hc3": [
-0.8985,
0.2139
]
},
"slope_wedge_weighted_wild_p": 0.1676,
"kish_n_weighted": 4.18,
"mean_dlog_gdp": 0.0905,
"mean_wedge": 0.1,
"median_wedge": 0.0889,
"intercept_demeaned": {
"b": 0.1,
"se": 0.0158,
"p": 0,
"stars": "***"
},
"intercept_raw": {
"b": 0.1551,
"se": 0.048,
"p": 0.0012,
"stars": "***"
},
"weighted_intercept_demeaned": {
"b": 0.0576,
"se": 0.0099,
"p": 0,
"stars": "***"
},
"aggregate_balanced": {
"n": 28,
"w_start": 0.1274,
"w_end": 0.1856,
"change": 0.0582
},
"upper_ci_times_mean_growth": 0.0167,
"growth_share_point": -0.5511,
"growth_share_ci": [
-1.2695,
0.1672
],
"mde_80pct_5pct": 1.1346
}
},
"core_no_norway_members": [
"Australia",
"Austria",
"Belgium",
"Canada",
"Cyprus",
"Denmark",
"Finland",
"France",
"Germany",
"Greece",
"Hong Kong",
"Ireland",
"Israel",
"Italy",
"Japan",
"Luxembourg",
"Malta",
"Netherlands",
"New Zealand",
"Portugal",
"Singapore",
"South Korea",
"Spain",
"Sweden",
"Switzerland",
"Taiwan",
"United Kingdom",
"United States"
],
"norway_co2_share_pct": {
"2007": 0.3,
"2017": 0.3,
"2022": 0.3
},
"oil_transition_norway_co2_share_pct": {
"1990": 6,
"1997": 5.8,
"2007": 7.9,
"2017": 10.8,
"2022": 11.4
},
"net_importers_1990": {
"members": [
"Austria",
"Canada",
"Cyprus",
"Denmark",
"Finland",
"France",
"Germany",
"Greece",
"Hong Kong",
"Ireland",
"Israel",
"Italy",
"Japan",
"Latvia",
"Lithuania",
"Malta",
"Netherlands",
"New Zealand",
"Portugal",
"Singapore",
"Slovakia",
"Slovenia",
"South Korea",
"Spain",
"Sweden",
"Switzerland",
"United Arab Emirates",
"United Kingdom"
],
"n": 28,
"n_with_1990_row": 38,
"net_exporters_1990": [
"Australia",
"Belgium",
"Czechia",
"Estonia",
"Luxembourg",
"Saudi Arabia",
"Taiwan",
"Trinidad and Tobago",
"United States",
"Venezuela"
],
"in_1990_not_1997_set": [
"Cyprus",
"Slovakia"
],
"in_1997_not_1990_set": [
"Taiwan"
],
"dropped_co2_share_pct": {
"1990": 52.9,
"2007": 55.9
},
"groups": {
"1990-2007": {
"n_high": 28,
"n_total": 106,
"slope_co2_hc1": {
"b": 0.6214,
"se": 0.1545,
"p": 0.0001,
"stars": "***",
"ci95": [
0.3186,
0.9242
]
},
"slope_consumption_co2_hc1": {
"b": 0.7262,
"se": 0.089,
"p": 0,
"stars": "***",
"ci95": [
0.5518,
0.9006
]
},
"slope_wedge_hc1": {
"b": 0.1049,
"se": 0.1197,
"p": 0.3811,
"stars": "",
"ci95": [
-0.1297,
0.3395
]
},
"slope_wedge_hc3": {
"b": 0.1049,
"se_hc3": 0.1301,
"p_hc3": 0.4204,
"ci95_hc3": [
-0.1502,
0.3599
]
},
"slope_wedge_wild_p": 0.3221,
"slope_wedge_weighted_hc1": {
"b": 0.1222,
"se": 0.0982,
"p": 0.2131,
"stars": ""
},
"slope_wedge_weighted_hc3": {
"b": 0.1222,
"se_hc3": 0.1186,
"p_hc3": 0.3026,
"ci95_hc3": [
-0.1102,
0.3547
]
},
"slope_wedge_weighted_wild_p": 0.2925,
"kish_n_weighted": 8.54,
"mean_dlog_gdp": 0.6257,
"mean_wedge": 0.1213,
"median_wedge": 0.0104,
"intercept_demeaned": {
"b": 0.1213,
"se": 0.0444,
"p": 0.0063,
"stars": "***"
},
"intercept_raw": {
"b": 0.0557,
"se": 0.0887,
"p": 0.5299,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0502,
"se": 0.0277,
"p": 0.07,
"stars": "*"
},
"aggregate_balanced": {
"n": 28,
"w_start": 0.1804,
"w_end": 0.2325,
"change": 0.0521
},
"upper_ci_times_mean_growth": 0.2124,
"growth_share_point": 0.5411,
"growth_share_ci": [
-0.669,
1.7512
],
"mde_80pct_5pct": 0.3354
},
"1997-2007": {
"n_high": 28,
"n_total": 107,
"slope_co2_hc1": {
"b": 0.3653,
"se": 0.171,
"p": 0.0327,
"stars": "**",
"ci95": [
0.0301,
0.7005
]
},
"slope_consumption_co2_hc1": {
"b": 0.6588,
"se": 0.0876,
"p": 0,
"stars": "***",
"ci95": [
0.4871,
0.8305
]
},
"slope_wedge_hc1": {
"b": 0.2935,
"se": 0.1595,
"p": 0.0658,
"stars": "*",
"ci95": [
-0.0191,
0.6061
]
},
"slope_wedge_hc3": {
"b": 0.2935,
"se_hc3": 0.2906,
"p_hc3": 0.3125,
"ci95_hc3": [
-0.2761,
0.8631
]
},
"slope_wedge_wild_p": 0.0971,
"slope_wedge_weighted_hc1": {
"b": 0.3772,
"se": 0.1462,
"p": 0.0099,
"stars": "***"
},
"slope_wedge_weighted_hc3": {
"b": 0.3772,
"se_hc3": 0.2478,
"p_hc3": 0.128,
"ci95_hc3": [
-0.1085,
0.863
]
},
"slope_wedge_weighted_wild_p": 0.1658,
"kish_n_weighted": 9.12,
"mean_dlog_gdp": 0.4184,
"mean_wedge": 0.1474,
"median_wedge": 0.1388,
"intercept_demeaned": {
"b": 0.1474,
"se": 0.0333,
"p": 0,
"stars": "***"
},
"intercept_raw": {
"b": 0.0246,
"se": 0.0713,
"p": 0.7299,
"stars": ""
},
"weighted_intercept_demeaned": {
"b": 0.0601,
"se": 0.0203,
"p": 0.0031,
"stars": "***"
},
"aggregate_balanced": {
"n": 28,
"w_start": 0.169,
"w_end": 0.2325,
"change": 0.0635
},
"upper_ci_times_mean_growth": 0.2536,
"growth_share_point": 0.8331,
"growth_share_ci": [
-0.0542,
1.7204
],
"mde_80pct_5pct": 0.4469
}
}
},
"inference_rule_wild_10pct": {
"rows": {
"all": {
"1990-2007": {
"b": 0.0274,
"p_hc1": 0.8155,
"p_hc3": 0.8421,
"p_wild": 0.8086,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": 0.0665,
"w_p_hc1": 0.519,
"w_p_wild": 0.5634,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"2007-2022": {
"b": 0.1866,
"p_hc1": 0.3805,
"p_hc3": 0.5048,
"p_wild": 0.5005,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": -0.0432,
"w_p_hc1": 0.1926,
"w_p_wild": 0.1495,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"1997-2007": {
"b": 0.0574,
"p_hc1": 0.6325,
"p_hc3": 0.6479,
"p_wild": 0.6125,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": 0.1369,
"w_p_hc1": 0.3235,
"w_p_wild": 0.5036,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"2007-2017": {
"b": 0.437,
"p_hc1": 0.2042,
"p_hc3": 0.2486,
"p_wild": 0.2127,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": 0.0804,
"w_p_hc1": 0.373,
"w_p_wild": 0.4562,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"2017-2022": {
"b": -0.0962,
"p_hc1": 0.2784,
"p_hc3": 0.7322,
"p_wild": 0.5581,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": -0.1945,
"w_p_hc1": 0,
"w_p_wild": 0.1892,
"w_sig10_hc1": true,
"w_sig10_wild": false
}
},
"no_oil_transition": {
"1990-2007": {
"b": 0.3055,
"p_hc1": 0.1912,
"p_hc3": 0.2998,
"p_wild": 0.3367,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": 0.2034,
"w_p_hc1": 0.226,
"w_p_wild": 0.3613,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"2007-2022": {
"b": 0.7383,
"p_hc1": 0.18,
"p_hc3": 0.2758,
"p_wild": 0.2132,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": -0.0522,
"w_p_hc1": 0.5273,
"w_p_wild": 0.5297,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"1997-2007": {
"b": 0.325,
"p_hc1": 0.3424,
"p_hc3": 0.3898,
"p_wild": 0.3646,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": 0.3771,
"w_p_hc1": 0.0934,
"w_p_wild": 0.468,
"w_sig10_hc1": true,
"w_sig10_wild": false
},
"2007-2017": {
"b": 0.9852,
"p_hc1": 0.1373,
"p_hc3": 0.2181,
"p_wild": 0.0675,
"sig10_hc1": false,
"sig10_wild": true,
"w_b": 0.0519,
"w_p_hc1": 0.7499,
"w_p_wild": 0.7802,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"2017-2022": {
"b": -0.5608,
"p_hc1": 0.183,
"p_hc3": 0.2549,
"p_wild": 0.2547,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": -0.342,
"w_p_hc1": 0.0439,
"w_p_wild": 0.1395,
"w_sig10_hc1": true,
"w_sig10_wild": false
}
},
"net_importers_1997": {
"1990-2007": {
"b": 0.1303,
"p_hc1": 0.2715,
"p_hc3": 0.344,
"p_wild": 0.1831,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": 0.1716,
"w_p_hc1": 0.1028,
"w_p_wild": 0.1759,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"2007-2022": {
"b": 0.527,
"p_hc1": 0.2913,
"p_hc3": 0.3783,
"p_wild": 0.4849,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": -0.0894,
"w_p_hc1": 0.3574,
"w_p_wild": 0.4546,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"1997-2007": {
"b": 0.2786,
"p_hc1": 0.0628,
"p_hc3": 0.2845,
"p_wild": 0.0908,
"sig10_hc1": true,
"sig10_wild": true,
"w_b": 0.3078,
"w_p_hc1": 0.0295,
"w_p_wild": 0.2224,
"w_sig10_hc1": true,
"w_sig10_wild": false
},
"2007-2017": {
"b": 0.6256,
"p_hc1": 0.2459,
"p_hc3": 0.3406,
"p_wild": 0.3462,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": -0.0648,
"w_p_hc1": 0.5661,
"w_p_wild": 0.5773,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"2017-2022": {
"b": -0.4263,
"p_hc1": 0.3082,
"p_hc3": 0.3846,
"p_wild": 0.5308,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": -0.0468,
"w_p_hc1": 0.773,
"w_p_wild": 0.7751,
"w_sig10_hc1": false,
"w_sig10_wild": false
}
},
"core_no_norway": {
"1990-2007": {
"b": 0.3055,
"p_hc1": 0.1912,
"p_hc3": 0.2998,
"p_wild": 0.3367,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": 0.2034,
"w_p_hc1": 0.226,
"w_p_wild": 0.3613,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"2007-2022": {
"b": 0.7809,
"p_hc1": 0.1499,
"p_hc3": 0.2437,
"p_wild": 0.1213,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": -0.0377,
"w_p_hc1": 0.6454,
"w_p_wild": 0.6449,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"1997-2007": {
"b": 0.325,
"p_hc1": 0.3424,
"p_hc3": 0.3898,
"p_wild": 0.3646,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": 0.3771,
"w_p_hc1": 0.0934,
"w_p_wild": 0.468,
"w_sig10_hc1": true,
"w_sig10_wild": false
},
"2007-2017": {
"b": 1.0243,
"p_hc1": 0.1268,
"p_hc3": 0.2092,
"p_wild": 0.0514,
"sig10_hc1": false,
"sig10_wild": true,
"w_b": 0.0622,
"w_p_hc1": 0.7062,
"w_p_wild": 0.7463,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"2017-2022": {
"b": -0.609,
"p_hc1": 0.1326,
"p_hc3": 0.1988,
"p_wild": 0.1571,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": -0.3423,
"w_p_hc1": 0.0498,
"w_p_wild": 0.1676,
"w_sig10_hc1": true,
"w_sig10_wild": false
}
},
"net_importers_1990": {
"1990-2007": {
"b": 0.1049,
"p_hc1": 0.3811,
"p_hc3": 0.4204,
"p_wild": 0.3221,
"sig10_hc1": false,
"sig10_wild": false,
"w_b": 0.1222,
"w_p_hc1": 0.2131,
"w_p_wild": 0.2925,
"w_sig10_hc1": false,
"w_sig10_wild": false
},
"1997-2007": {
"b": 0.2935,
"p_hc1": 0.0658,
"p_hc3": 0.3125,
"p_wild": 0.0971,
"sig10_hc1": true,
"sig10_wild": true,
"w_b": 0.3772,
"w_p_hc1": 0.0099,
"w_p_wild": 0.1658,
"w_sig10_hc1": true,
"w_sig10_wild": false
}
}
},
"sig_under_wild": [
"core_no_norway|2007-2017",
"net_importers_1990|1997-2007",
"net_importers_1997|1997-2007",
"no_oil_transition|2007-2017"
],
"sig_under_wild_weighted": [],
"status_differs_hc1_vs_wild": [
"core_no_norway|2007-2017",
"no_oil_transition|2007-2017"
],
"weighted_status_differs_hc1_vs_wild": [
"all|2017-2022",
"core_no_norway|1997-2007",
"core_no_norway|2017-2022",
"net_importers_1990|1997-2007",
"net_importers_1997|1997-2007",
"no_oil_transition|1997-2007",
"no_oil_transition|2017-2022"
]
}
},
"round6": {
"leave_one_out_passing": {
"core_no_norway|2007-2017": {
"n": 28,
"b_full": 1.0243,
"p_wild_full": 0.0514,
"seed": 20261159,
"loo_min": 0.3387,
"loo_min_dropped": "Malta",
"loo_max": 1.3156,
"loo_max_dropped": "Singapore",
"most_influential": "Malta",
"b_without_most_influential": 0.3387,
"p_hc1_without_most_influential": 0.0431,
"p_hc3_without_most_influential": 0.0656,
"p_wild_without_most_influential": 0.0436,
"loo_p_wild_range": [
0.0354,
0.1089
],
"n_loo_fits_sig10_wild": 26,
"n_loo_fits": 28,
"diagnostics": {
"max_leverage": {
"country": "Greece",
"value": 0.3123,
"mean_leverage": 0.0714
},
"max_abs_studentized_resid": {
"country": "Malta",
"value": 4.3702
},
"max_cooks_d": {
"country": "Malta",
"value": 2.0989
},
"max_share_hc_variance": {
"country": "Malta",
"value": 0.7438
},
"per_country": {
"Australia": {
"dlog_gdp": 0.2735,
"dw": 0.0154,
"leverage": 0.058,
"studentized_resid": -0.5161,
"cooks_d": 0.0082,
"share_hc_variance": 0.0018
},
"Austria": {
"dlog_gdp": 0.1049,
"dw": -0.0823,
"leverage": 0.0401,
"studentized_resid": -0.2488,
"cooks_d": 0.0013,
"share_hc_variance": 0.0001
},
"Belgium": {
"dlog_gdp": 0.1157,
"dw": 0.0817,
"leverage": 0.0385,
"studentized_resid": 0.2856,
"cooks_d": 0.0016,
"share_hc_variance": 0.0001
},
"Canada": {
"dlog_gdp": 0.1677,
"dw": -0.0679,
"leverage": 0.0359,
"studentized_resid": -0.4226,
"cooks_d": 0.0033,
"share_hc_variance": 0
},
"Cyprus": {
"dlog_gdp": 0.1052,
"dw": 0.1024,
"leverage": 0.0401,
"studentized_resid": 0.3958,
"cooks_d": 0.0033,
"share_hc_variance": 0.0002
},
"Denmark": {
"dlog_gdp": 0.1099,
"dw": 0.1906,
"leverage": 0.0393,
"studentized_resid": 0.6876,
"cooks_d": 0.0097,
"share_hc_variance": 0.0005
},
"Finland": {
"dlog_gdp": 0.0165,
"dw": -0.0076,
"leverage": 0.068,
"studentized_resid": 0.3339,
"cooks_d": 0.0041,
"share_hc_variance": 0.0011
},
"France": {
"dlog_gdp": 0.0793,
"dw": -0.0262,
"leverage": 0.0456,
"studentized_resid": 0.0391,
"cooks_d": 0,
"share_hc_variance": 0
},
"Germany": {
"dlog_gdp": 0.1677,
"dw": -0.0625,
"leverage": 0.0359,
"studentized_resid": -0.4039,
"cooks_d": 0.003,
"share_hc_variance": 0
},
"Greece": {
"dlog_gdp": -0.2538,
"dw": -0.0861,
"leverage": 0.3123,
"studentized_resid": 1.2084,
"cooks_d": 0.3316,
"share_hc_variance": 0.0913
},
"Hong Kong": {
"dlog_gdp": 0.2324,
"dw": 0.2558,
"leverage": 0.0451,
"studentized_resid": 0.4781,
"cooks_d": 0.0054,
"share_hc_variance": 0.0007
},
"Ireland": {
"dlog_gdp": 0.1458,
"dw": -0.2338,
"leverage": 0.0359,
"studentized_resid": -0.9234,
"cooks_d": 0.0159,
"share_hc_variance": 0.0001
},
"Israel": {
"dlog_gdp": 0.3518,
"dw": 0.0922,
"leverage": 0.0981,
"studentized_resid": -0.5401,
"cooks_d": 0.0159,
"share_hc_variance": 0.0054
},
"Italy": {
"dlog_gdp": -0.0228,
"dw": 0.0378,
"leverage": 0.0886,
"studentized_resid": 0.6453,
"cooks_d": 0.0202,
"share_hc_variance": 0.0066
},
"Japan": {
"dlog_gdp": 0.0491,
"dw": -0.0501,
"leverage": 0.0547,
"studentized_resid": 0.0639,
"cooks_d": 0.0001,
"share_hc_variance": 0
},
"Luxembourg": {
"dlog_gdp": 0.1578,
"dw": 0.0102,
"leverage": 0.0357,
"studentized_resid": -0.1149,
"cooks_d": 0.0002,
"share_hc_variance": 0
},
"Malta": {
"dlog_gdp": 0.4536,
"dw": 1.5007,
"leverage": 0.1802,
"studentized_resid": 4.3702,
"cooks_d": 2.0989,
"share_hc_variance": 0.7438
},
"Netherlands": {
"dlog_gdp": 0.1042,
"dw": -0.045,
"leverage": 0.0403,
"studentized_resid": -0.1159,
"cooks_d": 0.0003,
"share_hc_variance": 0
},
"New Zealand": {
"dlog_gdp": 0.2456,
"dw": -0.0395,
"leverage": 0.0486,
"studentized_resid": -0.606,
"cooks_d": 0.0094,
"share_hc_variance": 0.0015
},
"Portugal": {
"dlog_gdp": 0.0172,
"dw": -0.5709,
"leverage": 0.0677,
"studentized_resid": -1.668,
"cooks_d": 0.101,
"share_hc_variance": 0.0273
},
"Singapore": {
"dlog_gdp": 0.5337,
"dw": 0.0835,
"leverage": 0.2687,
"studentized_resid": -1.381,
"cooks_d": 0.3504,
"share_hc_variance": 0.1069
},
"South Korea": {
"dlog_gdp": 0.2807,
"dw": -0.1378,
"leverage": 0.0609,
"studentized_resid": -1.0851,
"cooks_d": 0.0382,
"share_hc_variance": 0.0092
},
"Spain": {
"dlog_gdp": 0.0639,
"dw": -0.0945,
"leverage": 0.0499,
"studentized_resid": -0.1455,
"cooks_d": 0.0006,
"share_hc_variance": 0.0001
},
"Sweden": {
"dlog_gdp": 0.1467,
"dw": -0.0037,
"leverage": 0.0359,
"studentized_resid": -0.1234,
"cooks_d": 0.0003,
"share_hc_variance": 0
},
"Switzerland": {
"dlog_gdp": 0.2169,
"dw": 0.2717,
"leverage": 0.0416,
"studentized_resid": 0.5885,
"cooks_d": 0.0075,
"share_hc_variance": 0.0006
},
"Taiwan": {
"dlog_gdp": 0.2763,
"dw": -0.0067,
"leverage": 0.0591,
"studentized_resid": -0.6047,
"cooks_d": 0.0115,
"share_hc_variance": 0.0026
},
"United Kingdom": {
"dlog_gdp": 0.1133,
"dw": 0.0529,
"leverage": 0.0388,
"studentized_resid": 0.1939,
"cooks_d": 0.0008,
"share_hc_variance": 0
},
"United States": {
"dlog_gdp": 0.1382,
"dw": -0.0024,
"leverage": 0.0363,
"studentized_resid": -0.0887,
"cooks_d": 0.0001,
"share_hc_variance": 0
}
}
},
"loo": {
"Australia": {
"b": 1.054,
"p_hc1": 0.1282,
"p_hc3": 0.2165,
"p_wild": 0.054
},
"Austria": {
"b": 1.018,
"p_hc1": 0.1284,
"p_hc3": 0.2115,
"p_wild": 0.0619
},
"Belgium": {
"b": 1.03,
"p_hc1": 0.1252,
"p_hc3": 0.2081,
"p_wild": 0.0502
},
"Canada": {
"b": 1.0265,
"p_hc1": 0.1256,
"p_hc3": 0.2087,
"p_wild": 0.0536
},
"Cyprus": {
"b": 1.0343,
"p_hc1": 0.1243,
"p_hc3": 0.2072,
"p_wild": 0.0489
},
"Denmark": {
"b": 1.04,
"p_hc1": 0.1228,
"p_hc3": 0.2057,
"p_wild": 0.0455
},
"Finland": {
"b": 1.0475,
"p_hc1": 0.1276,
"p_hc3": 0.2139,
"p_wild": 0.0601
},
"France": {
"b": 1.0258,
"p_hc1": 0.1277,
"p_hc3": 0.2112,
"p_wild": 0.0597
},
"Germany": {
"b": 1.0264,
"p_hc1": 0.1257,
"p_hc3": 0.2088,
"p_wild": 0.0545
},
"Greece": {
"b": 1.3107,
"p_hc1": 0.1236,
"p_hc3": 0.228,
"p_wild": 0.0488
},
"Hong Kong": {
"b": 1.0066,
"p_hc1": 0.1407,
"p_hc3": 0.2278,
"p_wild": 0.0673
},
"Ireland": {
"b": 1.0193,
"p_hc1": 0.1263,
"p_hc3": 0.2088,
"p_wild": 0.0599
},
"Israel": {
"b": 1.0774,
"p_hc1": 0.1391,
"p_hc3": 0.2363,
"p_wild": 0.0697
},
"Italy": {
"b": 1.0824,
"p_hc1": 0.123,
"p_hc3": 0.2123,
"p_wild": 0.053
},
"Japan": {
"b": 1.0277,
"p_hc1": 0.1297,
"p_hc3": 0.2143,
"p_wild": 0.0625
},
"Luxembourg": {
"b": 1.0243,
"p_hc1": 0.1266,
"p_hc3": 0.2097,
"p_wild": 0.0545
},
"Malta": {
"b": 0.3387,
"p_hc1": 0.0431,
"p_hc3": 0.0656,
"p_wild": 0.0436
},
"Netherlands": {
"b": 1.0213,
"p_hc1": 0.1276,
"p_hc3": 0.2106,
"p_wild": 0.0543
},
"New Zealand": {
"b": 1.0507,
"p_hc1": 0.1245,
"p_hc3": 0.2105,
"p_wild": 0.0441
},
"Portugal": {
"b": 0.9088,
"p_hc1": 0.1699,
"p_hc3": 0.2589,
"p_wild": 0.1023
},
"Singapore": {
"b": 1.3156,
"p_hc1": 0.1181,
"p_hc3": 0.2374,
"p_wild": 0.1089
},
"South Korea": {
"b": 1.0907,
"p_hc1": 0.1152,
"p_hc3": 0.2027,
"p_wild": 0.0354
},
"Spain": {
"b": 1.0177,
"p_hc1": 0.1312,
"p_hc3": 0.2154,
"p_wild": 0.0627
},
"Sweden": {
"b": 1.0237,
"p_hc1": 0.1265,
"p_hc3": 0.2094,
"p_wild": 0.0504
},
"Switzerland": {
"b": 1.007,
"p_hc1": 0.1388,
"p_hc3": 0.2251,
"p_wild": 0.0625
},
"Taiwan": {
"b": 1.0599,
"p_hc1": 0.1264,
"p_hc3": 0.2147,
"p_wild": 0.0496
},
"United Kingdom": {
"b": 1.0284,
"p_hc1": 0.1256,
"p_hc3": 0.2085,
"p_wild": 0.0505
},
"United States": {
"b": 1.0235,
"p_hc1": 0.1265,
"p_hc3": 0.2094,
"p_wild": 0.0533
}
},
"malta_in_sample": true,
"without_malta": {
"b": 0.3387,
"p_hc1": 0.0431,
"p_hc3": 0.0656,
"p_wild": 0.0436
}
},
"net_importers_1990|1997-2007": {
"n": 28,
"b_full": 0.2935,
"p_wild_full": 0.0971,
"seed": 20270211,
"loo_min": 0.2225,
"loo_min_dropped": "Singapore",
"loo_max": 0.5617,
"loo_max_dropped": "United Arab Emirates",
"most_influential": "United Arab Emirates",
"b_without_most_influential": 0.5617,
"p_hc1_without_most_influential": 0.0153,
"p_hc3_without_most_influential": 0.0275,
"p_wild_without_most_influential": 0.036,
"loo_p_wild_range": [
0.036,
0.1603
],
"n_loo_fits_sig10_wild": 5,
"n_loo_fits": 28,
"diagnostics": {
"max_leverage": {
"country": "United Arab Emirates",
"value": 0.5744,
"mean_leverage": 0.0714
},
"max_abs_studentized_resid": {
"country": "Singapore",
"value": 2.9985
},
"max_cooks_d": {
"country": "United Arab Emirates",
"value": 1.5764
},
"max_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.5426
},
"per_country": {
"Austria": {
"dlog_gdp": 0.2972,
"dw": -0.0037,
"leverage": 0.0471,
"studentized_resid": -0.6663,
"cooks_d": 0.011,
"share_hc_variance": 0.0049
},
"Canada": {
"dlog_gdp": 0.3405,
"dw": -0.0085,
"leverage": 0.0404,
"studentized_resid": -0.7642,
"cooks_d": 0.0123,
"share_hc_variance": 0.0027
},
"Cyprus": {
"dlog_gdp": 0.5121,
"dw": 0.2223,
"leverage": 0.0425,
"studentized_resid": 0.2726,
"cooks_d": 0.0016,
"share_hc_variance": 0.0005
},
"Denmark": {
"dlog_gdp": 0.2581,
"dw": 0.2283,
"leverage": 0.0555,
"studentized_resid": 0.741,
"cooks_d": 0.0161,
"share_hc_variance": 0.0104
},
"Finland": {
"dlog_gdp": 0.3767,
"dw": 0.1974,
"leverage": 0.0371,
"studentized_resid": 0.3572,
"cooks_d": 0.0025,
"share_hc_variance": 0.0002
},
"France": {
"dlog_gdp": 0.2595,
"dw": 0.1303,
"leverage": 0.0552,
"studentized_resid": 0.1709,
"cooks_d": 0.0009,
"share_hc_variance": 0.0005
},
"Germany": {
"dlog_gdp": 0.2689,
"dw": 0.0556,
"leverage": 0.053,
"studentized_resid": -0.277,
"cooks_d": 0.0021,
"share_hc_variance": 0.0013
},
"Greece": {
"dlog_gdp": 0.4743,
"dw": 0.0586,
"leverage": 0.0381,
"studentized_resid": -0.604,
"cooks_d": 0.0072,
"share_hc_variance": 0.0009
},
"Hong Kong": {
"dlog_gdp": 0.2987,
"dw": -0.2953,
"leverage": 0.0468,
"studentized_resid": -2.3499,
"cooks_d": 0.1354,
"share_hc_variance": 0.0589
},
"Ireland": {
"dlog_gdp": 0.7575,
"dw": 0.2608,
"leverage": 0.1244,
"studentized_resid": 0.0836,
"cooks_d": 0.0005,
"share_hc_variance": 0.0005
},
"Israel": {
"dlog_gdp": 0.3441,
"dw": -0.0032,
"leverage": 0.04,
"studentized_resid": -0.74,
"cooks_d": 0.0114,
"share_hc_variance": 0.0023
},
"Italy": {
"dlog_gdp": 0.2245,
"dw": 0.061,
"leverage": 0.0647,
"studentized_resid": -0.1716,
"cooks_d": 0.001,
"share_hc_variance": 0.0008
},
"Japan": {
"dlog_gdp": 0.0984,
"dw": -0.0658,
"leverage": 0.1147,
"studentized_resid": -0.7138,
"cooks_d": 0.033,
"share_hc_variance": 0.0361
},
"Latvia": {
"dlog_gdp": 0.6667,
"dw": 0.3864,
"leverage": 0.0832,
"studentized_resid": 0.9763,
"cooks_d": 0.0433,
"share_hc_variance": 0.0421
},
"Lithuania": {
"dlog_gdp": 0.6405,
"dw": 0.1474,
"leverage": 0.0737,
"studentized_resid": -0.3816,
"cooks_d": 0.0058,
"share_hc_variance": 0.0052
},
"Malta": {
"dlog_gdp": 0.325,
"dw": 0.3758,
"leverage": 0.0424,
"studentized_resid": 1.4714,
"cooks_d": 0.048,
"share_hc_variance": 0.0141
},
"Netherlands": {
"dlog_gdp": 0.3301,
"dw": 0.0038,
"leverage": 0.0417,
"studentized_resid": -0.6764,
"cooks_d": 0.01,
"share_hc_variance": 0.0027
},
"New Zealand": {
"dlog_gdp": 0.3586,
"dw": 0.0932,
"leverage": 0.0385,
"studentized_resid": -0.2105,
"cooks_d": 0.0009,
"share_hc_variance": 0.0001
},
"Portugal": {
"dlog_gdp": 0.2444,
"dw": 0.4332,
"leverage": 0.0591,
"studentized_resid": 1.9544,
"cooks_d": 0.1199,
"share_hc_variance": 0.085
},
"Singapore": {
"dlog_gdp": 0.5864,
"dw": 0.7139,
"leverage": 0.0575,
"studentized_resid": 2.9985,
"cooks_d": 0.274,
"share_hc_variance": 0.1867
},
"Slovakia": {
"dlog_gdp": 0.4617,
"dw": 0.1736,
"leverage": 0.0372,
"studentized_resid": 0.077,
"cooks_d": 0.0001,
"share_hc_variance": 0
},
"Slovenia": {
"dlog_gdp": 0.3845,
"dw": 0.0149,
"leverage": 0.0366,
"studentized_resid": -0.7028,
"cooks_d": 0.0094,
"share_hc_variance": 0.0004
},
"South Korea": {
"dlog_gdp": 0.4022,
"dw": 0.0918,
"leverage": 0.0359,
"studentized_resid": -0.2916,
"cooks_d": 0.0016,
"share_hc_variance": 0
},
"Spain": {
"dlog_gdp": 0.4698,
"dw": 0.1004,
"leverage": 0.0377,
"studentized_resid": -0.3564,
"cooks_d": 0.0025,
"share_hc_variance": 0.0003
},
"Sweden": {
"dlog_gdp": 0.3841,
"dw": 0.1672,
"leverage": 0.0366,
"studentized_resid": 0.1712,
"cooks_d": 0.0006,
"share_hc_variance": 0
},
"Switzerland": {
"dlog_gdp": 0.4173,
"dw": 0.2217,
"leverage": 0.0357,
"studentized_resid": 0.4278,
"cooks_d": 0.0034,
"share_hc_variance": 0
},
"United Arab Emirates": {
"dlog_gdp": 1.2544,
"dw": 0.2156,
"leverage": 0.5744,
"studentized_resid": -1.5285,
"cooks_d": 1.5764,
"share_hc_variance": 0.5426
},
"United Kingdom": {
"dlog_gdp": 0.28,
"dw": 0.1512,
"leverage": 0.0505,
"studentized_resid": 0.2563,
"cooks_d": 0.0017,
"share_hc_variance": 0.0009
}
}
},
"loo": {
"Austria": {
"b": 0.2822,
"p_hc1": 0.0746,
"p_hc3": 0.3248,
"p_wild": 0.0992
},
"Canada": {
"b": 0.2852,
"p_hc1": 0.0727,
"p_hc3": 0.3229,
"p_wild": 0.0977
},
"Cyprus": {
"b": 0.2899,
"p_hc1": 0.068,
"p_hc3": 0.3221,
"p_wild": 0.1014
},
"Denmark": {
"b": 0.3102,
"p_hc1": 0.0615,
"p_hc3": 0.3177,
"p_wild": 0.1074
},
"Finland": {
"b": 0.2956,
"p_hc1": 0.0638,
"p_hc3": 0.308,
"p_wild": 0.1011
},
"France": {
"b": 0.2973,
"p_hc1": 0.0688,
"p_hc3": 0.3229,
"p_wild": 0.105
},
"Germany": {
"b": 0.2877,
"p_hc1": 0.0731,
"p_hc3": 0.3251,
"p_wild": 0.1032
},
"Greece": {
"b": 0.2982,
"p_hc1": 0.0683,
"p_hc3": 0.33,
"p_wild": 0.1048
},
"Hong Kong": {
"b": 0.2541,
"p_hc1": 0.0845,
"p_hc3": 0.3397,
"p_wild": 0.082
},
"Ireland": {
"b": 0.2894,
"p_hc1": 0.0982,
"p_hc3": 0.441,
"p_wild": 0.1603
},
"Israel": {
"b": 0.2858,
"p_hc1": 0.0723,
"p_hc3": 0.3225,
"p_wild": 0.1035
},
"Italy": {
"b": 0.2888,
"p_hc1": 0.0758,
"p_hc3": 0.3337,
"p_wild": 0.1132
},
"Japan": {
"b": 0.2603,
"p_hc1": 0.099,
"p_hc3": 0.3789,
"p_wild": 0.1237
},
"Latvia": {
"b": 0.2588,
"p_hc1": 0.0805,
"p_hc3": 0.3505,
"p_wild": 0.132
},
"Lithuania": {
"b": 0.3056,
"p_hc1": 0.0764,
"p_hc3": 0.3802,
"p_wild": 0.1265
},
"Malta": {
"b": 0.3127,
"p_hc1": 0.0543,
"p_hc3": 0.2982,
"p_wild": 0.1035
},
"Netherlands": {
"b": 0.2851,
"p_hc1": 0.0726,
"p_hc3": 0.3223,
"p_wild": 0.0996
},
"New Zealand": {
"b": 0.2917,
"p_hc1": 0.0678,
"p_hc3": 0.3156,
"p_wild": 0.1002
},
"Portugal": {
"b": 0.3415,
"p_hc1": 0.043,
"p_hc3": 0.308,
"p_wild": 0.1179
},
"Singapore": {
"b": 0.2225,
"p_hc1": 0.0433,
"p_hc3": 0.2307,
"p_wild": 0.1589
},
"Slovakia": {
"b": 0.293,
"p_hc1": 0.0663,
"p_hc3": 0.317,
"p_wild": 0.1062
},
"Slovenia": {
"b": 0.2902,
"p_hc1": 0.0702,
"p_hc3": 0.3221,
"p_wild": 0.1029
},
"South Korea": {
"b": 0.2929,
"p_hc1": 0.0674,
"p_hc3": 0.3172,
"p_wild": 0.1044
},
"Spain": {
"b": 0.2961,
"p_hc1": 0.0676,
"p_hc3": 0.3252,
"p_wild": 0.1078
},
"Sweden": {
"b": 0.2943,
"p_hc1": 0.0649,
"p_hc3": 0.3104,
"p_wild": 0.1056
},
"Switzerland": {
"b": 0.2936,
"p_hc1": 0.0639,
"p_hc3": 0.3076,
"p_wild": 0.1045
},
"United Arab Emirates": {
"b": 0.5617,
"p_hc1": 0.0153,
"p_hc3": 0.0275,
"p_wild": 0.036
},
"United Kingdom": {
"b": 0.2985,
"p_hc1": 0.0667,
"p_hc3": 0.3182,
"p_wild": 0.1016
}
},
"malta_in_sample": true,
"without_malta": {
"b": 0.3127,
"p_hc1": 0.0543,
"p_hc3": 0.2982,
"p_wild": 0.1035
}
},
"net_importers_1997|1997-2007": {
"n": 27,
"b_full": 0.2786,
"p_wild_full": 0.0908,
"seed": 20261249,
"loo_min": 0.2025,
"loo_min_dropped": "Singapore",
"loo_max": 0.5116,
"loo_max_dropped": "United Arab Emirates",
"most_influential": "United Arab Emirates",
"b_without_most_influential": 0.5116,
"p_hc1_without_most_influential": 0.0386,
"p_hc3_without_most_influential": 0.059,
"p_wild_without_most_influential": 0.0687,
"loo_p_wild_range": [
0.0687,
0.1444
],
"n_loo_fits_sig10_wild": 15,
"n_loo_fits": 27,
"diagnostics": {
"max_leverage": {
"country": "United Arab Emirates",
"value": 0.5849,
"mean_leverage": 0.0741
},
"max_abs_studentized_resid": {
"country": "Singapore",
"value": 2.8257
},
"max_cooks_d": {
"country": "United Arab Emirates",
"value": 0.9732
},
"max_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.4421
},
"per_country": {
"Austria": {
"dlog_gdp": 0.2972,
"dw": -0.0037,
"leverage": 0.0477,
"studentized_resid": -0.5281,
"cooks_d": 0.007,
"share_hc_variance": 0.004
},
"Canada": {
"dlog_gdp": 0.3405,
"dw": -0.0085,
"leverage": 0.0413,
"studentized_resid": -0.6139,
"cooks_d": 0.0081,
"share_hc_variance": 0.0022
},
"Denmark": {
"dlog_gdp": 0.2581,
"dw": 0.2283,
"leverage": 0.056,
"studentized_resid": 0.7471,
"cooks_d": 0.0166,
"share_hc_variance": 0.0141
},
"Finland": {
"dlog_gdp": 0.3767,
"dw": 0.1974,
"leverage": 0.0382,
"studentized_resid": 0.4072,
"cooks_d": 0.0033,
"share_hc_variance": 0.0003
},
"France": {
"dlog_gdp": 0.2595,
"dw": 0.1303,
"leverage": 0.0557,
"studentized_resid": 0.2296,
"cooks_d": 0.0016,
"share_hc_variance": 0.0013
},
"Germany": {
"dlog_gdp": 0.2689,
"dw": 0.0556,
"leverage": 0.0535,
"studentized_resid": -0.1765,
"cooks_d": 0.0009,
"share_hc_variance": 0.0007
},
"Greece": {
"dlog_gdp": 0.4743,
"dw": 0.0586,
"leverage": 0.0398,
"studentized_resid": -0.4583,
"cooks_d": 0.0044,
"share_hc_variance": 0.0008
},
"Hong Kong": {
"dlog_gdp": 0.2987,
"dw": -0.2953,
"leverage": 0.0475,
"studentized_resid": -2.0568,
"cooks_d": 0.1054,
"share_hc_variance": 0.0591
},
"Ireland": {
"dlog_gdp": 0.7575,
"dw": 0.2608,
"leverage": 0.1284,
"studentized_resid": 0.1941,
"cooks_d": 0.0028,
"share_hc_variance": 0.0042
},
"Israel": {
"dlog_gdp": 0.3441,
"dw": -0.0032,
"leverage": 0.0409,
"studentized_resid": -0.5916,
"cooks_d": 0.0075,
"share_hc_variance": 0.0018
},
"Italy": {
"dlog_gdp": 0.2245,
"dw": 0.061,
"leverage": 0.0651,
"studentized_resid": -0.0838,
"cooks_d": 0.0002,
"share_hc_variance": 0.0003
},
"Japan": {
"dlog_gdp": 0.0984,
"dw": -0.0658,
"leverage": 0.1147,
"studentized_resid": -0.5842,
"cooks_d": 0.0221,
"share_hc_variance": 0.033
},
"Latvia": {
"dlog_gdp": 0.6667,
"dw": 0.3864,
"leverage": 0.0864,
"studentized_resid": 0.9958,
"cooks_d": 0.0469,
"share_hc_variance": 0.063
},
"Lithuania": {
"dlog_gdp": 0.6405,
"dw": 0.1474,
"leverage": 0.0767,
"studentized_resid": -0.2415,
"cooks_d": 0.0024,
"share_hc_variance": 0.003
},
"Malta": {
"dlog_gdp": 0.325,
"dw": 0.3758,
"leverage": 0.0433,
"studentized_resid": 1.4151,
"cooks_d": 0.0453,
"share_hc_variance": 0.0168
},
"Netherlands": {
"dlog_gdp": 0.3301,
"dw": 0.0038,
"leverage": 0.0426,
"studentized_resid": -0.5349,
"cooks_d": 0.0064,
"share_hc_variance": 0.0021
},
"New Zealand": {
"dlog_gdp": 0.3586,
"dw": 0.0932,
"leverage": 0.0395,
"studentized_resid": -0.1098,
"cooks_d": 0.0002,
"share_hc_variance": 0
},
"Portugal": {
"dlog_gdp": 0.2444,
"dw": 0.4332,
"leverage": 0.0595,
"studentized_resid": 1.8478,
"cooks_d": 0.1081,
"share_hc_variance": 0.1017
},
"Singapore": {
"dlog_gdp": 0.5864,
"dw": 0.7139,
"leverage": 0.06,
"studentized_resid": 2.8257,
"cooks_d": 0.2547,
"share_hc_variance": 0.2422
},
"Slovenia": {
"dlog_gdp": 0.3845,
"dw": 0.0149,
"leverage": 0.0377,
"studentized_resid": -0.555,
"cooks_d": 0.006,
"share_hc_variance": 0.0003
},
"South Korea": {
"dlog_gdp": 0.4022,
"dw": 0.0918,
"leverage": 0.0372,
"studentized_resid": -0.1801,
"cooks_d": 0.0006,
"share_hc_variance": 0
},
"Spain": {
"dlog_gdp": 0.4698,
"dw": 0.1004,
"leverage": 0.0394,
"studentized_resid": -0.2337,
"cooks_d": 0.0011,
"share_hc_variance": 0.0002
},
"Sweden": {
"dlog_gdp": 0.3841,
"dw": 0.1672,
"leverage": 0.0378,
"studentized_resid": 0.2388,
"cooks_d": 0.0011,
"share_hc_variance": 0.0001
},
"Switzerland": {
"dlog_gdp": 0.4173,
"dw": 0.2217,
"leverage": 0.037,
"studentized_resid": 0.4745,
"cooks_d": 0.0043,
"share_hc_variance": 0
},
"Taiwan": {
"dlog_gdp": 0.4509,
"dw": -0.2265,
"leverage": 0.0381,
"studentized_resid": -1.909,
"cooks_d": 0.0721,
"share_hc_variance": 0.0051
},
"United Arab Emirates": {
"dlog_gdp": 1.2544,
"dw": 0.2156,
"leverage": 0.5849,
"studentized_resid": -1.1753,
"cooks_d": 0.9732,
"share_hc_variance": 0.4421
},
"United Kingdom": {
"dlog_gdp": 0.28,
"dw": 0.1512,
"leverage": 0.0511,
"studentized_resid": 0.3086,
"cooks_d": 0.0026,
"share_hc_variance": 0.0018
}
}
},
"loo": {
"Austria": {
"b": 0.269,
"p_hc1": 0.0705,
"p_hc3": 0.2949,
"p_wild": 0.0971
},
"Canada": {
"b": 0.2715,
"p_hc1": 0.0686,
"p_hc3": 0.2938,
"p_wild": 0.0973
},
"Denmark": {
"b": 0.2969,
"p_hc1": 0.0561,
"p_hc3": 0.2885,
"p_wild": 0.1082
},
"Finland": {
"b": 0.281,
"p_hc1": 0.0603,
"p_hc3": 0.2781,
"p_wild": 0.0976
},
"France": {
"b": 0.2842,
"p_hc1": 0.0639,
"p_hc3": 0.2932,
"p_wild": 0.1055
},
"Germany": {
"b": 0.2746,
"p_hc1": 0.0687,
"p_hc3": 0.2951,
"p_wild": 0.1015
},
"Greece": {
"b": 0.2829,
"p_hc1": 0.0646,
"p_hc3": 0.3029,
"p_wild": 0.1033
},
"Hong Kong": {
"b": 0.2415,
"p_hc1": 0.0811,
"p_hc3": 0.3086,
"p_wild": 0.0814
},
"Ireland": {
"b": 0.2678,
"p_hc1": 0.0991,
"p_hc3": 0.4116,
"p_wild": 0.1444
},
"Israel": {
"b": 0.2721,
"p_hc1": 0.0682,
"p_hc3": 0.2935,
"p_wild": 0.0963
},
"Italy": {
"b": 0.2761,
"p_hc1": 0.0709,
"p_hc3": 0.3035,
"p_wild": 0.1031
},
"Japan": {
"b": 0.2488,
"p_hc1": 0.0955,
"p_hc3": 0.3445,
"p_wild": 0.1137
},
"Latvia": {
"b": 0.2387,
"p_hc1": 0.0817,
"p_hc3": 0.3061,
"p_wild": 0.1051
},
"Lithuania": {
"b": 0.2872,
"p_hc1": 0.0742,
"p_hc3": 0.3548,
"p_wild": 0.1156
},
"Malta": {
"b": 0.2983,
"p_hc1": 0.0501,
"p_hc3": 0.2682,
"p_wild": 0.0996
},
"Netherlands": {
"b": 0.2716,
"p_hc1": 0.0685,
"p_hc3": 0.293,
"p_wild": 0.095
},
"New Zealand": {
"b": 0.2777,
"p_hc1": 0.0639,
"p_hc3": 0.2864,
"p_wild": 0.0969
},
"Portugal": {
"b": 0.3279,
"p_hc1": 0.0367,
"p_hc3": 0.2799,
"p_wild": 0.1136
},
"Singapore": {
"b": 0.2025,
"p_hc1": 0.045,
"p_hc3": 0.1621,
"p_wild": 0.1014
},
"Slovenia": {
"b": 0.276,
"p_hc1": 0.0662,
"p_hc3": 0.2938,
"p_wild": 0.097
},
"South Korea": {
"b": 0.2783,
"p_hc1": 0.0637,
"p_hc3": 0.2884,
"p_wild": 0.0965
},
"Spain": {
"b": 0.2806,
"p_hc1": 0.0642,
"p_hc3": 0.2972,
"p_wild": 0.0989
},
"Sweden": {
"b": 0.2797,
"p_hc1": 0.0614,
"p_hc3": 0.2808,
"p_wild": 0.0964
},
"Switzerland": {
"b": 0.2784,
"p_hc1": 0.0609,
"p_hc3": 0.2773,
"p_wild": 0.0943
},
"Taiwan": {
"b": 0.2894,
"p_hc1": 0.0685,
"p_hc3": 0.3271,
"p_wild": 0.105
},
"United Arab Emirates": {
"b": 0.5116,
"p_hc1": 0.0386,
"p_hc3": 0.059,
"p_wild": 0.0687
},
"United Kingdom": {
"b": 0.2851,
"p_hc1": 0.062,
"p_hc3": 0.2886,
"p_wild": 0.0957
}
},
"malta_in_sample": true,
"without_malta": {
"b": 0.2983,
"p_hc1": 0.0501,
"p_hc3": 0.2682,
"p_wild": 0.0996
}
},
"no_oil_transition|2007-2017": {
"n": 29,
"b_full": 0.9852,
"p_wild_full": 0.0675,
"seed": 20261159,
"loo_min": 0.3039,
"loo_min_dropped": "Malta",
"loo_max": 1.2654,
"loo_max_dropped": "Greece",
"most_influential": "Malta",
"b_without_most_influential": 0.3039,
"p_hc1_without_most_influential": 0.0686,
"p_hc3_without_most_influential": 0.0912,
"p_wild_without_most_influential": 0.0684,
"loo_p_wild_range": [
0.0474,
0.1475
],
"n_loo_fits_sig10_wild": 27,
"n_loo_fits": 29,
"diagnostics": {
"max_leverage": {
"country": "Greece",
"value": 0.312,
"mean_leverage": 0.069
},
"max_abs_studentized_resid": {
"country": "Malta",
"value": 4.4313
},
"max_cooks_d": {
"country": "Malta",
"value": 2.0815
},
"max_share_hc_variance": {
"country": "Malta",
"value": 0.7609
},
"per_country": {
"Australia": {
"dlog_gdp": 0.2735,
"dw": 0.0154,
"leverage": 0.0556,
"studentized_resid": -0.4605,
"cooks_d": 0.0062,
"share_hc_variance": 0.0014
},
"Austria": {
"dlog_gdp": 0.1049,
"dw": -0.0823,
"leverage": 0.0393,
"studentized_resid": -0.2177,
"cooks_d": 0.001,
"share_hc_variance": 0.0001
},
"Belgium": {
"dlog_gdp": 0.1157,
"dw": 0.0817,
"leverage": 0.0376,
"studentized_resid": 0.3168,
"cooks_d": 0.002,
"share_hc_variance": 0.0001
},
"Canada": {
"dlog_gdp": 0.1677,
"dw": -0.0679,
"leverage": 0.0346,
"studentized_resid": -0.3825,
"cooks_d": 0.0026,
"share_hc_variance": 0
},
"Cyprus": {
"dlog_gdp": 0.1052,
"dw": 0.1024,
"leverage": 0.0393,
"studentized_resid": 0.4253,
"cooks_d": 0.0037,
"share_hc_variance": 0.0003
},
"Denmark": {
"dlog_gdp": 0.1099,
"dw": 0.1906,
"leverage": 0.0385,
"studentized_resid": 0.7169,
"cooks_d": 0.0103,
"share_hc_variance": 0.0007
},
"Finland": {
"dlog_gdp": 0.0165,
"dw": -0.0076,
"leverage": 0.0677,
"studentized_resid": 0.3518,
"cooks_d": 0.0045,
"share_hc_variance": 0.0013
},
"France": {
"dlog_gdp": 0.0793,
"dw": -0.0262,
"leverage": 0.0449,
"studentized_resid": 0.0661,
"cooks_d": 0.0001,
"share_hc_variance": 0
},
"Germany": {
"dlog_gdp": 0.1677,
"dw": -0.0625,
"leverage": 0.0346,
"studentized_resid": -0.3638,
"cooks_d": 0.0024,
"share_hc_variance": 0
},
"Greece": {
"dlog_gdp": -0.2538,
"dw": -0.0861,
"leverage": 0.312,
"studentized_resid": 1.1838,
"cooks_d": 0.3178,
"share_hc_variance": 0.0895
},
"Hong Kong": {
"dlog_gdp": 0.2324,
"dw": 0.2558,
"leverage": 0.0431,
"studentized_resid": 0.5245,
"cooks_d": 0.0062,
"share_hc_variance": 0.0008
},
"Ireland": {
"dlog_gdp": 0.1458,
"dw": -0.2338,
"leverage": 0.0348,
"studentized_resid": -0.885,
"cooks_d": 0.0141,
"share_hc_variance": 0.0001
},
"Israel": {
"dlog_gdp": 0.3518,
"dw": 0.0922,
"leverage": 0.0945,
"studentized_resid": -0.4719,
"cooks_d": 0.0116,
"share_hc_variance": 0.004
},
"Italy": {
"dlog_gdp": -0.0228,
"dw": 0.0378,
"leverage": 0.0885,
"studentized_resid": 0.6572,
"cooks_d": 0.021,
"share_hc_variance": 0.0071
},
"Japan": {
"dlog_gdp": 0.0491,
"dw": -0.0501,
"leverage": 0.0543,
"studentized_resid": 0.0868,
"cooks_d": 0.0002,
"share_hc_variance": 0
},
"Luxembourg": {
"dlog_gdp": 0.1578,
"dw": 0.0102,
"leverage": 0.0345,
"studentized_resid": -0.0771,
"cooks_d": 0.0001,
"share_hc_variance": 0
},
"Malta": {
"dlog_gdp": 0.4536,
"dw": 1.5007,
"leverage": 0.1749,
"studentized_resid": 4.4313,
"cooks_d": 2.0815,
"share_hc_variance": 0.7609
},
"Netherlands": {
"dlog_gdp": 0.1042,
"dw": -0.045,
"leverage": 0.0395,
"studentized_resid": -0.0853,
"cooks_d": 0.0001,
"share_hc_variance": 0
},
"New Zealand": {
"dlog_gdp": 0.2456,
"dw": -0.0395,
"leverage": 0.0465,
"studentized_resid": -0.5543,
"cooks_d": 0.0075,
"share_hc_variance": 0.0012
},
"Norway": {
"dlog_gdp": 0.2359,
"dw": -0.1918,
"leverage": 0.044,
"studentized_resid": -1.0529,
"cooks_d": 0.0255,
"share_hc_variance": 0.0034
},
"Portugal": {
"dlog_gdp": 0.0172,
"dw": -0.5709,
"leverage": 0.0674,
"studentized_resid": -1.6455,
"cooks_d": 0.0978,
"share_hc_variance": 0.0278
},
"Singapore": {
"dlog_gdp": 0.5337,
"dw": 0.0835,
"leverage": 0.2619,
"studentized_resid": -1.2703,
"cooks_d": 0.2863,
"share_hc_variance": 0.0906
},
"South Korea": {
"dlog_gdp": 0.2807,
"dw": -0.1378,
"leverage": 0.0583,
"studentized_resid": -1.0264,
"cooks_d": 0.0326,
"share_hc_variance": 0.0079
},
"Spain": {
"dlog_gdp": 0.0639,
"dw": -0.0945,
"leverage": 0.0493,
"studentized_resid": -0.1201,
"cooks_d": 0.0004,
"share_hc_variance": 0.0001
},
"Sweden": {
"dlog_gdp": 0.1467,
"dw": -0.0037,
"leverage": 0.0348,
"studentized_resid": -0.087,
"cooks_d": 0.0001,
"share_hc_variance": 0
},
"Switzerland": {
"dlog_gdp": 0.2169,
"dw": 0.2717,
"leverage": 0.0398,
"studentized_resid": 0.6324,
"cooks_d": 0.0083,
"share_hc_variance": 0.0007
},
"Taiwan": {
"dlog_gdp": 0.2763,
"dw": -0.0067,
"leverage": 0.0566,
"studentized_resid": -0.5484,
"cooks_d": 0.009,
"share_hc_variance": 0.0021
},
"United Kingdom": {
"dlog_gdp": 0.1133,
"dw": 0.0529,
"leverage": 0.038,
"studentized_resid": 0.225,
"cooks_d": 0.001,
"share_hc_variance": 0.0001
},
"United States": {
"dlog_gdp": 0.1382,
"dw": -0.0024,
"leverage": 0.0352,
"studentized_resid": -0.0535,
"cooks_d": 0.0001,
"share_hc_variance": 0
}
}
},
"loo": {
"Australia": {
"b": 1.0109,
"p_hc1": 0.1393,
"p_hc3": 0.2253,
"p_wild": 0.0676
},
"Austria": {
"b": 0.9795,
"p_hc1": 0.1391,
"p_hc3": 0.2204,
"p_wild": 0.0724
},
"Belgium": {
"b": 0.9919,
"p_hc1": 0.1354,
"p_hc3": 0.2166,
"p_wild": 0.0621
},
"Canada": {
"b": 0.9867,
"p_hc1": 0.1363,
"p_hc3": 0.2175,
"p_wild": 0.0675
},
"Cyprus": {
"b": 0.9964,
"p_hc1": 0.1343,
"p_hc3": 0.2155,
"p_wild": 0.0601
},
"Denmark": {
"b": 1.0025,
"p_hc1": 0.1325,
"p_hc3": 0.2138,
"p_wild": 0.0553
},
"Finland": {
"b": 1.01,
"p_hc1": 0.1376,
"p_hc3": 0.2222,
"p_wild": 0.0681
},
"France": {
"b": 0.9878,
"p_hc1": 0.1381,
"p_hc3": 0.2199,
"p_wild": 0.0686
},
"Germany": {
"b": 0.9867,
"p_hc1": 0.1364,
"p_hc3": 0.2176,
"p_wild": 0.0652
},
"Greece": {
"b": 1.2654,
"p_hc1": 0.1324,
"p_hc3": 0.2329,
"p_wild": 0.0698
},
"Hong Kong": {
"b": 0.9667,
"p_hc1": 0.152,
"p_hc3": 0.2369,
"p_wild": 0.084
},
"Ireland": {
"b": 0.9793,
"p_hc1": 0.1373,
"p_hc3": 0.218,
"p_wild": 0.0722
},
"Israel": {
"b": 1.0305,
"p_hc1": 0.1511,
"p_hc3": 0.2453,
"p_wild": 0.0824
},
"Italy": {
"b": 1.0449,
"p_hc1": 0.1322,
"p_hc3": 0.2199,
"p_wild": 0.0618
},
"Japan": {
"b": 0.9899,
"p_hc1": 0.1401,
"p_hc3": 0.2231,
"p_wild": 0.075
},
"Luxembourg": {
"b": 0.9852,
"p_hc1": 0.1372,
"p_hc3": 0.2185,
"p_wild": 0.0668
},
"Malta": {
"b": 0.3039,
"p_hc1": 0.0686,
"p_hc3": 0.0912,
"p_wild": 0.0684
},
"Netherlands": {
"b": 0.9829,
"p_hc1": 0.1382,
"p_hc3": 0.2195,
"p_wild": 0.0738
},
"New Zealand": {
"b": 1.0084,
"p_hc1": 0.1353,
"p_hc3": 0.2192,
"p_wild": 0.0643
},
"Norway": {
"b": 1.0243,
"p_hc1": 0.1268,
"p_hc3": 0.2092,
"p_wild": 0.0514
},
"Portugal": {
"b": 0.87,
"p_hc1": 0.1846,
"p_hc3": 0.2711,
"p_wild": 0.1475
},
"Singapore": {
"b": 1.248,
"p_hc1": 0.135,
"p_hc3": 0.2527,
"p_wild": 0.1425
},
"South Korea": {
"b": 1.0461,
"p_hc1": 0.1257,
"p_hc3": 0.211,
"p_wild": 0.0474
},
"Spain": {
"b": 0.9796,
"p_hc1": 0.142,
"p_hc3": 0.2244,
"p_wild": 0.0797
},
"Sweden": {
"b": 0.9847,
"p_hc1": 0.1371,
"p_hc3": 0.2182,
"p_wild": 0.0662
},
"Switzerland": {
"b": 0.9677,
"p_hc1": 0.15,
"p_hc3": 0.2341,
"p_wild": 0.0778
},
"Taiwan": {
"b": 1.0165,
"p_hc1": 0.1374,
"p_hc3": 0.2234,
"p_wild": 0.0666
},
"United Kingdom": {
"b": 0.9903,
"p_hc1": 0.1359,
"p_hc3": 0.2171,
"p_wild": 0.0654
},
"United States": {
"b": 0.9847,
"p_hc1": 0.1371,
"p_hc3": 0.2182,
"p_wild": 0.069
}
},
"malta_in_sample": true,
"without_malta": {
"b": 0.3039,
"p_hc1": 0.0686,
"p_hc3": 0.0912,
"p_wild": 0.0684
}
}
},
"malta_drives_core_2007_2017": {
"no_oil_transition|2007-2017": true,
"core_no_norway|2007-2017": true
},
"table6_endpoints_wild": {
"avg3_1990-92_to_2005-07": {
"wedge": {
"b": 0.0464,
"se_hc1": 0.2491,
"p_hc1": 0.8522,
"p_hc3": 0.8693,
"p_wild": 0.8823,
"sig10_hc1": false,
"sig10_wild": false
},
"wedge_weighted": {
"b": 0.093,
"se_hc1": 0.1584,
"p_hc1": 0.557,
"p_hc3": 0.6629,
"p_wild": 0.618,
"sig10_hc1": false,
"sig10_wild": false
},
"seed": 20270301
},
"benchmark_1997-2007": {
"wedge": {
"b": 0.0574,
"se_hc1": 0.12,
"p_hc1": 0.6325,
"p_hc3": 0.6479,
"p_wild": 0.6125,
"sig10_hc1": false,
"sig10_wild": false
},
"wedge_weighted": {
"b": 0.1369,
"se_hc1": 0.1387,
"p_hc1": 0.3235,
"p_hc3": 0.4519,
"p_wild": 0.5036,
"sig10_hc1": false,
"sig10_wild": false
},
"seed": 20261049
},
"avg3_2005-07_to_2020-22": {
"wedge": {
"b": 0.0962,
"se_hc1": 0.2093,
"p_hc1": 0.6458,
"p_hc3": 0.7096,
"p_wild": 0.876,
"sig10_hc1": false,
"sig10_wild": false
},
"wedge_weighted": {
"b": -0.0756,
"se_hc1": 0.0572,
"p_hc1": 0.1862,
"p_hc3": 0.2894,
"p_wild": 0.3544,
"sig10_hc1": false,
"sig10_wild": false
},
"seed": 20270321
},
"benchmark_2007-2017": {
"wedge": {
"b": 0.437,
"se_hc1": 0.3443,
"p_hc1": 0.2042,
"p_hc3": 0.2486,
"p_wild": 0.2127,
"sig10_hc1": false,
"sig10_wild": false
},
"wedge_weighted": {
"b": 0.0804,
"se_hc1": 0.0902,
"p_hc1": 0.373,
"p_hc3": 0.4665,
"p_wild": 0.4562,
"sig10_hc1": false,
"sig10_wild": false
},
"seed": 20261059
},
"end2019_2007-2019": {
"wedge": {
"b": 0.5355,
"se_hc1": 0.2745,
"p_hc1": 0.0511,
"p_hc3": 0.081,
"p_wild": 0.2719,
"sig10_hc1": true,
"sig10_wild": false
},
"wedge_weighted": {
"b": 0.238,
"se_hc1": 0.1747,
"p_hc1": 0.1729,
"p_hc3": 0.3986,
"p_wild": 0.6013,
"sig10_hc1": false,
"sig10_wild": false
},
"seed": 20270341
}
},
"table5_long_difference_wild": {
"start1995_pre": {
"b": 0.1009,
"se_hc1": 0.0957,
"p_hc1": 0.2915,
"p_hc3": 0.3302,
"p_wild": 0.3243,
"sig10_hc1": false,
"sig10_wild": false,
"n_high": 39
},
"pop1m_pre": {
"b": 0.0343,
"se_hc1": 0.1175,
"p_hc1": 0.7705,
"p_hc3": 0.8017,
"p_wild": 0.7665,
"sig10_hc1": false,
"sig10_wild": false,
"n_high": 35
},
"pop1m_post": {
"b": -0.0087,
"se_hc1": 0.0535,
"p_hc1": 0.8711,
"p_hc3": 0.8798,
"p_wild": 0.859,
"sig10_hc1": false,
"sig10_wild": false,
"n_high": 37
},
"percapita_pre": {
"b": 0.1456,
"se_hc1": 0.2095,
"p_hc1": 0.487,
"p_hc3": 0.5135,
"p_wild": 0.4949,
"sig10_hc1": false,
"sig10_wild": false,
"n_high": 38
},
"percapita_post": {
"b": 0.208,
"se_hc1": 0.2087,
"p_hc1": 0.3188,
"p_hc3": 0.5264,
"p_wild": 0.4211,
"sig10_hc1": false,
"sig10_wild": false,
"n_high": 40
},
"kuwait_keep_all_pre": {
"b": 0.002,
"se_hc1": 0.0884,
"p_hc1": 0.9817,
"p_hc3": 0.9838,
"p_wild": 0.9804,
"sig10_hc1": false,
"sig10_wild": false,
"n_high": 39
},
"kuwait_drop1991_92_pre": {
"b": 0.002,
"se_hc1": 0.0884,
"p_hc1": 0.9817,
"p_hc3": 0.9838,
"p_wild": 0.9804,
"sig10_hc1": false,
"sig10_wild": false,
"n_high": 39
}
},
"multiple_testing": {
"table4_unweighted": {
"n_tests": 20,
"n_pass_10pct": 4,
"expected_pass_under_null_10pct": 2,
"prob_ge_observed_if_independent": 0.133,
"min_p": 0.0514,
"bonferroni_threshold_10pct": 0.005,
"bonferroni_threshold_5pct": 0.0025,
"bonferroni_min_adjusted_p": 1,
"holm_min_adjusted_p": 1,
"n_reject_bonferroni_10pct": 0,
"n_reject_holm_10pct": 0,
"n_reject_holm_5pct": 0,
"labels": [
"all|1990-2007",
"all|2007-2022",
"all|1997-2007",
"all|2007-2017",
"all|2017-2022",
"no_oil_transition|1990-2007",
"no_oil_transition|2007-2022",
"no_oil_transition|1997-2007",
"no_oil_transition|2007-2017",
"no_oil_transition|2017-2022",
"net_importers_1997|1990-2007",
"net_importers_1997|2007-2022",
"net_importers_1997|1997-2007",
"net_importers_1997|2007-2017",
"net_importers_1997|2017-2022",
"core_no_norway|2007-2022",
"core_no_norway|2007-2017",
"core_no_norway|2017-2022",
"net_importers_1990|1990-2007",
"net_importers_1990|1997-2007"
]
},
"table4_weighted": {
"n_tests": 20,
"n_pass_10pct": 0,
"expected_pass_under_null_10pct": 2,
"prob_ge_observed_if_independent": 1,
"min_p": 0.1395,
"bonferroni_threshold_10pct": 0.005,
"bonferroni_threshold_5pct": 0.0025,
"bonferroni_min_adjusted_p": 1,
"holm_min_adjusted_p": 1,
"n_reject_bonferroni_10pct": 0,
"n_reject_holm_10pct": 0,
"n_reject_holm_5pct": 0,
"labels": [
"all|1990-2007",
"all|2007-2022",
"all|1997-2007",
"all|2007-2017",
"all|2017-2022",
"no_oil_transition|1990-2007",
"no_oil_transition|2007-2022",
"no_oil_transition|1997-2007",
"no_oil_transition|2007-2017",
"no_oil_transition|2017-2022",
"net_importers_1997|1990-2007",
"net_importers_1997|2007-2022",
"net_importers_1997|1997-2007",
"net_importers_1997|2007-2017",
"net_importers_1997|2017-2022",
"core_no_norway|2007-2022",
"core_no_norway|2007-2017",
"core_no_norway|2017-2022",
"net_importers_1990|1990-2007",
"net_importers_1990|1997-2007"
]
},
"all_long_difference_unweighted": {
"n_tests": 29,
"n_pass_10pct": 4,
"expected_pass_under_null_10pct": 2.9,
"prob_ge_observed_if_independent": 0.329,
"min_p": 0.0514,
"bonferroni_threshold_10pct": 0.0034,
"bonferroni_threshold_5pct": 0.0017,
"bonferroni_min_adjusted_p": 1,
"holm_min_adjusted_p": 1,
"n_reject_bonferroni_10pct": 0,
"n_reject_holm_10pct": 0,
"n_reject_holm_5pct": 0,
"labels": [
"all|1990-2007",
"all|2007-2022",
"all|1997-2007",
"all|2007-2017",
"all|2017-2022",
"no_oil_transition|1990-2007",
"no_oil_transition|2007-2022",
"no_oil_transition|1997-2007",
"no_oil_transition|2007-2017",
"no_oil_transition|2017-2022",
"net_importers_1997|1990-2007",
"net_importers_1997|2007-2022",
"net_importers_1997|1997-2007",
"net_importers_1997|2007-2017",
"net_importers_1997|2017-2022",
"core_no_norway|2007-2022",
"core_no_norway|2007-2017",
"core_no_norway|2017-2022",
"net_importers_1990|1990-2007",
"net_importers_1990|1997-2007",
"table6|avg3_1990-92_to_2005-07",
"table6|avg3_2005-07_to_2020-22",
"table6|end2019_2007-2019",
"table5|start1995_pre",
"table5|pop1m_pre",
"table5|pop1m_post",
"table5|percapita_pre",
"table5|percapita_post",
"table5|kuwait_keep_all_pre"
]
},
"note": "table4: distinct rows of round5 inference_rule_wild_10pct (core_no_norway 1990-2007 and 1997-2007 equal the core rows); all_long_difference adds table 6 avg3/end2019 rows and table 5 long-difference rows (Kuwait 1991-92 variant equals keep_all); tests are nested and overlapping, not independent"
},
"weighted_intercepts_net_importers": {
"net_importers_1997": {
"1990-2007": {
"b": 0.0594,
"se": 0.0282,
"p": 0.0351,
"stars": "**"
},
"2007-2017": {
"b": -0.0429,
"se": 0.0157,
"p": 0.0062,
"stars": "***"
}
},
"net_importers_1990": {
"1990-2007": {
"b": 0.0502,
"se": 0.0277,
"p": 0.07,
"stars": "*"
},
"1997-2007": {
"b": 0.0601,
"se": 0.0203,
"p": 0.0031,
"stars": "***"
}
}
}
},
"round7": {
"malta_facts": {
"1990": {
"co2_mt": 2.427,
"consumption_co2_mt": 3.263,
"wedge": 0.2957,
"population": 367913
},
"1997": {
"co2_mt": 2.556,
"consumption_co2_mt": 3.3,
"wedge": 0.2555,
"population": 390687
},
"2007": {
"co2_mt": 2.728,
"consumption_co2_mt": 5.128,
"wedge": 0.6313,
"population": 417201
},
"2017": {
"co2_mt": 1.537,
"consumption_co2_mt": 12.961,
"wedge": 2.132,
"population": 467902
},
"2019": {
"co2_mt": 1.649,
"consumption_co2_mt": 15.868,
"wedge": 2.264,
"population": 504017
},
"2022": {
"co2_mt": 1.774,
"consumption_co2_mt": 16.119,
"wedge": 2.2066,
"population": 528194
},
"tercile": "high"
},
"table4": {
"all": {
"1990-2007": {
"unweighted": {
"seed": 20261029,
"full": {
"n": 38,
"b": 0.0274,
"se_hc1": 0.1174,
"p_hc1": 0.8155,
"se_hc3": 0.1375,
"p_hc3": 0.8421,
"p_wild": 0.8086
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1053,
"dropped": [
"Latvia",
"Singapore",
"United Arab Emirates"
],
"n": 35,
"b": -0.1315,
"p_hc3": 0.4137,
"p_wild": 0.3778
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.4094
},
"without_malta": {
"n": 37,
"b": 0.0259,
"se_hc1": 0.119,
"p_hc1": 0.8278,
"se_hc3": 0.1409,
"p_hc3": 0.8542,
"p_wild": 0.8335
},
"malta": {
"leverage": 0.0266,
"mean_leverage": 0.0526,
"studentized_resid": 0.8836,
"cooks_d": 0.0107,
"share_hc_variance": 0.0002,
"rank_share_hc_variance": 27,
"rank_abs_studentized_resid": 10,
"rank_cooks_d": 17,
"dlog_gdp": 0.6769,
"dw": 0.3356
}
},
"weighted": {
"seed": 20261030,
"full": {
"n": 38,
"b": 0.0665,
"se_hc1": 0.1031,
"p_hc1": 0.519,
"se_hc3": 0.1254,
"p_hc3": 0.5959,
"p_wild": 0.5634
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1053,
"dropped": [
"Japan",
"Saudi Arabia",
"Singapore"
],
"n": 35,
"b": -0.0134,
"p_hc3": 0.9414,
"p_wild": 0.9237
},
"largest_share_hc_variance": {
"country": "Saudi Arabia",
"value": 0.2759
},
"without_malta": {
"n": 37,
"b": 0.0663,
"se_hc1": 0.1032,
"p_hc1": 0.5205,
"se_hc3": 0.1254,
"p_hc3": 0.597,
"p_wild": 0.5544
},
"malta": {
"leverage": 0.0003,
"mean_leverage": 0.0526,
"studentized_resid": 0.1891,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 34,
"rank_abs_studentized_resid": 31,
"rank_cooks_d": 37,
"dlog_gdp": 0.6769,
"dw": 0.3356,
"weight_share": 0.00021
}
}
},
"2007-2022": {
"unweighted": {
"seed": 20261039,
"full": {
"n": 40,
"b": 0.1866,
"se_hc1": 0.2128,
"p_hc1": 0.3805,
"se_hc3": 0.2798,
"p_hc3": 0.5048,
"p_wild": 0.5005
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Malta",
"Venezuela"
],
"n": 38,
"b": -0.0158,
"p_hc3": 0.906,
"p_wild": 0.9039
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.7381
},
"without_malta": {
"n": 39,
"b": -0.0075,
"se_hc1": 0.0532,
"p_hc1": 0.8885,
"se_hc3": 0.0572,
"p_hc3": 0.8961,
"p_wild": 0.8847
},
"malta": {
"leverage": 0.082,
"mean_leverage": 0.05,
"studentized_resid": 4.7776,
"cooks_d": 1.0198,
"share_hc_variance": 0.7381,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.6692,
"dw": 1.5753
}
},
"weighted": {
"seed": 20261040,
"full": {
"n": 40,
"b": -0.0432,
"se_hc1": 0.0332,
"p_hc1": 0.1926,
"se_hc3": 0.0385,
"p_hc3": 0.2615,
"p_wild": 0.1495
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [],
"n": 40,
"b": -0.0432,
"p_hc3": 0.2615,
"p_wild": 0.1495
},
"largest_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.3626
},
"without_malta": {
"n": 39,
"b": -0.0465,
"se_hc1": 0.0336,
"p_hc1": 0.1659,
"se_hc3": 0.04,
"p_hc3": 0.2442,
"p_wild": 0.1343
},
"malta": {
"leverage": 0.0012,
"mean_leverage": 0.05,
"studentized_resid": 1.569,
"cooks_d": 0.0014,
"share_hc_variance": 0.0104,
"rank_share_hc_variance": 14,
"rank_abs_studentized_resid": 6,
"rank_cooks_d": 25,
"dlog_gdp": 0.6692,
"dw": 1.5753,
"weight_share": 0.00019
}
}
},
"1997-2007": {
"unweighted": {
"seed": 20261049,
"full": {
"n": 39,
"b": 0.0574,
"se_hc1": 0.12,
"p_hc1": 0.6325,
"se_hc3": 0.1256,
"p_hc3": 0.6479,
"p_wild": 0.6125
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1026,
"dropped": [
"Belgium",
"Singapore",
"Trinidad and Tobago"
],
"n": 36,
"b": 0.129,
"p_hc3": 0.1523,
"p_wild": 0.1526
},
"largest_share_hc_variance": {
"country": "Trinidad and Tobago",
"value": 0.2424
},
"without_malta": {
"n": 38,
"b": 0.074,
"se_hc1": 0.1194,
"p_hc1": 0.5357,
"se_hc3": 0.1248,
"p_hc3": 0.5535,
"p_wild": 0.5159
},
"malta": {
"leverage": 0.0333,
"mean_leverage": 0.0513,
"studentized_resid": 1.262,
"cooks_d": 0.0274,
"share_hc_variance": 0.0188,
"rank_share_hc_variance": 11,
"rank_abs_studentized_resid": 7,
"rank_cooks_d": 10,
"dlog_gdp": 0.325,
"dw": 0.3758
}
},
"weighted": {
"seed": 20261050,
"full": {
"n": 39,
"b": 0.1369,
"se_hc1": 0.1387,
"p_hc1": 0.3235,
"se_hc3": 0.182,
"p_hc3": 0.4519,
"p_wild": 0.5036
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1026,
"dropped": [
"Japan",
"Saudi Arabia"
],
"n": 37,
"b": 0.0698,
"p_hc3": 0.5249,
"p_wild": 0.4641
},
"largest_share_hc_variance": {
"country": "Japan",
"value": 0.5587
},
"without_malta": {
"n": 38,
"b": 0.1369,
"se_hc1": 0.1387,
"p_hc1": 0.3237,
"se_hc3": 0.182,
"p_hc3": 0.4518,
"p_wild": 0.4994
},
"malta": {
"leverage": 0.0002,
"mean_leverage": 0.0513,
"studentized_resid": 0.2582,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 39,
"rank_abs_studentized_resid": 25,
"rank_cooks_d": 38,
"dlog_gdp": 0.325,
"dw": 0.3758,
"weight_share": 0.0002
}
}
},
"2007-2017": {
"unweighted": {
"seed": 20261059,
"full": {
"n": 40,
"b": 0.437,
"se_hc1": 0.3443,
"p_hc1": 0.2042,
"se_hc3": 0.3788,
"p_hc3": 0.2486,
"p_wild": 0.2127
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Malta",
"United Arab Emirates"
],
"n": 38,
"b": 0.2275,
"p_hc3": 0.0334,
"p_wild": 0.0371
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.7223
},
"without_malta": {
"n": 39,
"b": 0.1247,
"se_hc1": 0.1222,
"p_hc1": 0.3077,
"se_hc3": 0.1399,
"p_hc3": 0.3728,
"p_wild": 0.348
},
"malta": {
"leverage": 0.0873,
"mean_leverage": 0.05,
"studentized_resid": 5.1505,
"cooks_d": 1.2686,
"share_hc_variance": 0.7223,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.4536,
"dw": 1.5007
}
},
"weighted": {
"seed": 20261060,
"full": {
"n": 40,
"b": 0.0804,
"se_hc1": 0.0902,
"p_hc1": 0.373,
"se_hc3": 0.1104,
"p_hc3": 0.4665,
"p_wild": 0.4562
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Saudi Arabia",
"South Korea",
"United Arab Emirates",
"United States"
],
"n": 36,
"b": 0.1922,
"p_hc3": 0.0526,
"p_wild": 0.0951
},
"largest_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.325
},
"without_malta": {
"n": 39,
"b": 0.0752,
"se_hc1": 0.0908,
"p_hc1": 0.4075,
"se_hc3": 0.112,
"p_hc3": 0.502,
"p_wild": 0.4849
},
"malta": {
"leverage": 0.0013,
"mean_leverage": 0.05,
"studentized_resid": 1.7501,
"cooks_d": 0.0019,
"share_hc_variance": 0.0034,
"rank_share_hc_variance": 12,
"rank_abs_studentized_resid": 3,
"rank_cooks_d": 22,
"dlog_gdp": 0.4536,
"dw": 1.5007,
"weight_share": 0.00019
}
}
},
"2017-2022": {
"unweighted": {
"seed": 20261069,
"full": {
"n": 40,
"b": -0.0962,
"se_hc1": 0.0887,
"p_hc1": 0.2784,
"se_hc3": 0.2811,
"p_hc3": 0.7322,
"p_wild": 0.5581
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Hong Kong",
"Trinidad and Tobago",
"Venezuela"
],
"n": 37,
"b": 0.0723,
"p_hc3": 0.7256,
"p_wild": 0.7148
},
"largest_share_hc_variance": {
"country": "Trinidad and Tobago",
"value": 0.4835
},
"without_malta": {
"n": 39,
"b": -0.0976,
"se_hc1": 0.0891,
"p_hc1": 0.2733,
"se_hc3": 0.3067,
"p_hc3": 0.7503,
"p_wild": 0.5427
},
"malta": {
"leverage": 0.0487,
"mean_leverage": 0.05,
"studentized_resid": 0.068,
"cooks_d": 0.0001,
"share_hc_variance": 0.0002,
"rank_share_hc_variance": 24,
"rank_abs_studentized_resid": 35,
"rank_cooks_d": 34,
"dlog_gdp": 0.2156,
"dw": 0.0747
}
},
"weighted": {
"seed": 20261070,
"full": {
"n": 40,
"b": -0.1945,
"se_hc1": 0.0396,
"p_hc1": 0,
"se_hc3": 0.0885,
"p_hc3": 0.0279,
"p_wild": 0.1892
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Saudi Arabia",
"United States",
"Venezuela"
],
"n": 37,
"b": -0.0223,
"p_hc3": 0.9168,
"p_wild": 0.9086
},
"largest_share_hc_variance": {
"country": "Trinidad and Tobago",
"value": 0.185
},
"without_malta": {
"n": 39,
"b": -0.1946,
"se_hc1": 0.0396,
"p_hc1": 0,
"se_hc3": 0.0889,
"p_hc3": 0.0285,
"p_wild": 0.1852
},
"malta": {
"leverage": 0.0004,
"mean_leverage": 0.05,
"studentized_resid": 0.061,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 37,
"rank_abs_studentized_resid": 39,
"rank_cooks_d": 40,
"dlog_gdp": 0.2156,
"dw": 0.0747,
"weight_share": 0.00012
}
}
}
},
"no_oil_transition": {
"1990-2007": {
"unweighted": {
"seed": 20261129,
"full": {
"n": 28,
"b": 0.3055,
"se_hc1": 0.2337,
"p_hc1": 0.1912,
"se_hc3": 0.2946,
"p_hc3": 0.2998,
"p_wild": 0.3367
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Belgium",
"Singapore"
],
"n": 26,
"b": 0.145,
"p_hc3": 0.3962,
"p_wild": 0.4039
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.672
},
"without_malta": {
"n": 27,
"b": 0.2974,
"se_hc1": 0.2382,
"p_hc1": 0.2119,
"se_hc3": 0.3025,
"p_hc3": 0.3256,
"p_wild": 0.3639
},
"malta": {
"leverage": 0.0382,
"mean_leverage": 0.0714,
"studentized_resid": 0.8714,
"cooks_d": 0.0151,
"share_hc_variance": 0.0012,
"rank_share_hc_variance": 17,
"rank_abs_studentized_resid": 11,
"rank_cooks_d": 14,
"dlog_gdp": 0.6769,
"dw": 0.3356
}
},
"weighted": {
"seed": 20261130,
"full": {
"n": 28,
"b": 0.2034,
"se_hc1": 0.168,
"p_hc1": 0.226,
"se_hc3": 0.2097,
"p_hc3": 0.3323,
"p_wild": 0.3613
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Japan",
"Singapore",
"South Korea"
],
"n": 25,
"b": 0.0593,
"p_hc3": 0.8522,
"p_wild": 0.8747
},
"largest_share_hc_variance": {
"country": "South Korea",
"value": 0.3627
},
"without_malta": {
"n": 27,
"b": 0.2029,
"se_hc1": 0.1683,
"p_hc1": 0.2279,
"se_hc3": 0.2099,
"p_hc3": 0.3337,
"p_wild": 0.3477
},
"malta": {
"leverage": 0.0006,
"mean_leverage": 0.0714,
"studentized_resid": 0.1566,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 23,
"rank_abs_studentized_resid": 24,
"rank_cooks_d": 27,
"dlog_gdp": 0.6769,
"dw": 0.3356,
"weight_share": 0.00022
}
}
},
"2007-2022": {
"unweighted": {
"seed": 20261139,
"full": {
"n": 29,
"b": 0.7383,
"se_hc1": 0.5506,
"p_hc1": 0.18,
"se_hc3": 0.6775,
"p_hc3": 0.2758,
"p_wild": 0.2132
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1379,
"dropped": [
"Malta"
],
"n": 28,
"b": 0.1018,
"p_hc3": 0.5305,
"p_wild": 0.5048
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.867
},
"without_malta": {
"n": 28,
"b": 0.1018,
"se_hc1": 0.1427,
"p_hc1": 0.4758,
"se_hc3": 0.1623,
"p_hc3": 0.5305,
"p_wild": 0.5048
},
"malta": {
"leverage": 0.2172,
"mean_leverage": 0.069,
"studentized_resid": 3.8955,
"cooks_d": 2.1057,
"share_hc_variance": 0.867,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.6692,
"dw": 1.5753
}
},
"weighted": {
"seed": 20261140,
"full": {
"n": 29,
"b": -0.0522,
"se_hc1": 0.0826,
"p_hc1": 0.5273,
"se_hc3": 0.0945,
"p_hc3": 0.5807,
"p_wild": 0.5297
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1379,
"dropped": [],
"n": 29,
"b": -0.0522,
"p_hc3": 0.5807,
"p_wild": 0.5297
},
"largest_share_hc_variance": {
"country": "South Korea",
"value": 0.1948
},
"without_malta": {
"n": 28,
"b": -0.0645,
"se_hc1": 0.0826,
"p_hc1": 0.4349,
"se_hc3": 0.0953,
"p_hc3": 0.4988,
"p_wild": 0.4443
},
"malta": {
"leverage": 0.0038,
"mean_leverage": 0.069,
"studentized_resid": 1.3795,
"cooks_d": 0.0036,
"share_hc_variance": 0.0235,
"rank_share_hc_variance": 11,
"rank_abs_studentized_resid": 6,
"rank_cooks_d": 16,
"dlog_gdp": 0.6692,
"dw": 1.5753,
"weight_share": 0.00021
}
}
},
"1997-2007": {
"unweighted": {
"seed": 20261149,
"full": {
"n": 28,
"b": 0.325,
"se_hc1": 0.3423,
"p_hc1": 0.3424,
"se_hc3": 0.378,
"p_hc3": 0.3898,
"p_wild": 0.3646
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Belgium",
"Singapore"
],
"n": 26,
"b": 0.1738,
"p_hc3": 0.537,
"p_wild": 0.5076
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.5866
},
"without_malta": {
"n": 27,
"b": 0.3481,
"se_hc1": 0.344,
"p_hc1": 0.3115,
"se_hc3": 0.3795,
"p_hc3": 0.359,
"p_wild": 0.3307
},
"malta": {
"leverage": 0.0391,
"mean_leverage": 0.0714,
"studentized_resid": 1.2037,
"cooks_d": 0.0295,
"share_hc_variance": 0.0045,
"rank_share_hc_variance": 10,
"rank_abs_studentized_resid": 6,
"rank_cooks_d": 7,
"dlog_gdp": 0.325,
"dw": 0.3758
}
},
"weighted": {
"seed": 20261150,
"full": {
"n": 28,
"b": 0.3771,
"se_hc1": 0.2248,
"p_hc1": 0.0934,
"se_hc3": 0.403,
"p_hc3": 0.3495,
"p_wild": 0.468
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Japan",
"Taiwan"
],
"n": 26,
"b": 0.2221,
"p_hc3": 0.462,
"p_wild": 0.5274
},
"largest_share_hc_variance": {
"country": "Japan",
"value": 0.4717
},
"without_malta": {
"n": 27,
"b": 0.3768,
"se_hc1": 0.2251,
"p_hc1": 0.0941,
"se_hc3": 0.403,
"p_hc3": 0.3497,
"p_wild": 0.4629
},
"malta": {
"leverage": 0.0002,
"mean_leverage": 0.0714,
"studentized_resid": 0.2244,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 27,
"rank_abs_studentized_resid": 20,
"rank_cooks_d": 27,
"dlog_gdp": 0.325,
"dw": 0.3758,
"weight_share": 0.00021
}
}
},
"2007-2017": {
"unweighted": {
"seed": 20261159,
"full": {
"n": 29,
"b": 0.9852,
"se_hc1": 0.6631,
"p_hc1": 0.1373,
"se_hc3": 0.7999,
"p_hc3": 0.2181,
"p_wild": 0.0675
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1379,
"dropped": [
"Greece",
"Malta",
"Singapore"
],
"n": 26,
"b": 0.4329,
"p_hc3": 0.3131,
"p_wild": 0.3262
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.7609
},
"without_malta": {
"n": 28,
"b": 0.3039,
"se_hc1": 0.1669,
"p_hc1": 0.0686,
"se_hc3": 0.1799,
"p_hc3": 0.0912,
"p_wild": 0.0684
},
"malta": {
"leverage": 0.1749,
"mean_leverage": 0.069,
"studentized_resid": 4.4313,
"cooks_d": 2.0815,
"share_hc_variance": 0.7609,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.4536,
"dw": 1.5007
}
},
"weighted": {
"seed": 20261160,
"full": {
"n": 29,
"b": 0.0519,
"se_hc1": 0.1628,
"p_hc1": 0.7499,
"se_hc3": 0.1873,
"p_hc3": 0.7817,
"p_wild": 0.7802
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1379,
"dropped": [
"South Korea",
"United States"
],
"n": 27,
"b": 0.1742,
"p_hc3": 0.2149,
"p_wild": 0.1929
},
"largest_share_hc_variance": {
"country": "South Korea",
"value": 0.5708
},
"without_malta": {
"n": 28,
"b": 0.0357,
"se_hc1": 0.1613,
"p_hc1": 0.8247,
"se_hc3": 0.1857,
"p_hc3": 0.8474,
"p_wild": 0.8475
},
"malta": {
"leverage": 0.0037,
"mean_leverage": 0.069,
"studentized_resid": 1.6434,
"cooks_d": 0.005,
"share_hc_variance": 0.0105,
"rank_share_hc_variance": 11,
"rank_abs_studentized_resid": 3,
"rank_cooks_d": 16,
"dlog_gdp": 0.4536,
"dw": 1.5007,
"weight_share": 0.00021
}
}
},
"2017-2022": {
"unweighted": {
"seed": 20261169,
"full": {
"n": 29,
"b": -0.5608,
"se_hc1": 0.4211,
"p_hc1": 0.183,
"se_hc3": 0.4925,
"p_hc3": 0.2549,
"p_wild": 0.2547
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1379,
"dropped": [
"Hong Kong",
"Norway"
],
"n": 27,
"b": -0.1784,
"p_hc3": 0.2787,
"p_wild": 0.2621
},
"largest_share_hc_variance": {
"country": "Hong Kong",
"value": 0.8309
},
"without_malta": {
"n": 28,
"b": -0.6507,
"se_hc1": 0.4725,
"p_hc1": 0.1684,
"se_hc3": 0.5626,
"p_hc3": 0.2474,
"p_wild": 0.2548
},
"malta": {
"leverage": 0.1928,
"mean_leverage": 0.069,
"studentized_resid": 0.5908,
"cooks_d": 0.0417,
"share_hc_variance": 0.0315,
"rank_share_hc_variance": 3,
"rank_abs_studentized_resid": 11,
"rank_cooks_d": 4,
"dlog_gdp": 0.2156,
"dw": 0.0747
}
},
"weighted": {
"seed": 20261170,
"full": {
"n": 29,
"b": -0.342,
"se_hc1": 0.1697,
"p_hc1": 0.0439,
"se_hc3": 0.2671,
"p_hc3": 0.2004,
"p_wild": 0.1395
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1379,
"dropped": [
"Japan",
"United States"
],
"n": 27,
"b": -0.3066,
"p_hc3": 0.2565,
"p_wild": 0.1621
},
"largest_share_hc_variance": {
"country": "Japan",
"value": 0.2025
},
"without_malta": {
"n": 28,
"b": -0.3426,
"se_hc1": 0.1702,
"p_hc1": 0.0442,
"se_hc3": 0.2677,
"p_hc3": 0.2007,
"p_wild": 0.1436
},
"malta": {
"leverage": 0.0014,
"mean_leverage": 0.069,
"studentized_resid": 0.0813,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 25,
"rank_abs_studentized_resid": 28,
"rank_cooks_d": 28,
"dlog_gdp": 0.2156,
"dw": 0.0747,
"weight_share": 0.00014
}
}
}
},
"net_importers_1997": {
"1990-2007": {
"unweighted": {
"seed": 20261229,
"full": {
"n": 27,
"b": 0.1303,
"se_hc1": 0.1185,
"p_hc1": 0.2715,
"se_hc3": 0.1377,
"p_hc3": 0.344,
"p_wild": 0.1831
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"Latvia",
"Singapore"
],
"n": 25,
"b": 0.1213,
"p_hc3": 0.0298,
"p_wild": 0.0623
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.4545
},
"without_malta": {
"n": 26,
"b": 0.1282,
"se_hc1": 0.1192,
"p_hc1": 0.2822,
"se_hc3": 0.1361,
"p_hc3": 0.3461,
"p_wild": 0.2067
},
"malta": {
"leverage": 0.0375,
"mean_leverage": 0.0741,
"studentized_resid": 0.7818,
"cooks_d": 0.0119,
"share_hc_variance": 0.0003,
"rank_share_hc_variance": 20,
"rank_abs_studentized_resid": 10,
"rank_cooks_d": 12,
"dlog_gdp": 0.6769,
"dw": 0.3356
}
},
"weighted": {
"seed": 20261230,
"full": {
"n": 27,
"b": 0.1716,
"se_hc1": 0.1052,
"p_hc1": 0.1028,
"se_hc3": 0.1244,
"p_hc3": 0.1676,
"p_wild": 0.1759
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"Japan",
"Singapore",
"South Korea",
"United Kingdom"
],
"n": 23,
"b": 0.1474,
"p_hc3": 0.1285,
"p_wild": 0.0911
},
"largest_share_hc_variance": {
"country": "South Korea",
"value": 0.312
},
"without_malta": {
"n": 26,
"b": 0.1713,
"se_hc1": 0.1053,
"p_hc1": 0.104,
"se_hc3": 0.1243,
"p_hc3": 0.1682,
"p_wild": 0.1763
},
"malta": {
"leverage": 0.0007,
"mean_leverage": 0.0741,
"studentized_resid": 0.1879,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 22,
"rank_abs_studentized_resid": 25,
"rank_cooks_d": 26,
"dlog_gdp": 0.6769,
"dw": 0.3356,
"weight_share": 0.00044
}
}
},
"2007-2022": {
"unweighted": {
"seed": 20261239,
"full": {
"n": 27,
"b": 0.527,
"se_hc1": 0.4994,
"p_hc1": 0.2913,
"se_hc3": 0.5981,
"p_hc3": 0.3783,
"p_wild": 0.4849
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"Malta",
"United Arab Emirates"
],
"n": 25,
"b": 0.1861,
"p_hc3": 0.1493,
"p_wild": 0.1321
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.7363
},
"without_malta": {
"n": 26,
"b": 0.0208,
"se_hc1": 0.1615,
"p_hc1": 0.8977,
"se_hc3": 0.204,
"p_hc3": 0.9189,
"p_wild": 0.9176
},
"malta": {
"leverage": 0.1778,
"mean_leverage": 0.0741,
"studentized_resid": 3.9693,
"cooks_d": 1.7034,
"share_hc_variance": 0.7363,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.6692,
"dw": 1.5753
}
},
"weighted": {
"seed": 20261240,
"full": {
"n": 27,
"b": -0.0894,
"se_hc1": 0.0972,
"p_hc1": 0.3574,
"se_hc3": 0.1204,
"p_hc3": 0.4578,
"p_wild": 0.4546
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"Japan",
"United Arab Emirates"
],
"n": 25,
"b": -0.0576,
"p_hc3": 0.6697,
"p_wild": 0.6451
},
"largest_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.3891
},
"without_malta": {
"n": 26,
"b": -0.1016,
"se_hc1": 0.095,
"p_hc1": 0.2851,
"se_hc3": 0.1177,
"p_hc3": 0.3883,
"p_wild": 0.3973
},
"malta": {
"leverage": 0.0042,
"mean_leverage": 0.0741,
"studentized_resid": 1.431,
"cooks_d": 0.0043,
"share_hc_variance": 0.0168,
"rank_share_hc_variance": 10,
"rank_abs_studentized_resid": 5,
"rank_cooks_d": 18,
"dlog_gdp": 0.6692,
"dw": 1.5753,
"weight_share": 0.00043
}
}
},
"1997-2007": {
"unweighted": {
"seed": 20261249,
"full": {
"n": 27,
"b": 0.2786,
"se_hc1": 0.1497,
"p_hc1": 0.0628,
"se_hc3": 0.2603,
"p_hc3": 0.2845,
"p_wild": 0.0908
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"Singapore",
"United Arab Emirates"
],
"n": 25,
"b": 0.3111,
"p_hc3": 0.1428,
"p_wild": 0.1838
},
"largest_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.4421
},
"without_malta": {
"n": 26,
"b": 0.2983,
"se_hc1": 0.1523,
"p_hc1": 0.0501,
"se_hc3": 0.2694,
"p_hc3": 0.2682,
"p_wild": 0.0996
},
"malta": {
"leverage": 0.0433,
"mean_leverage": 0.0741,
"studentized_resid": 1.4151,
"cooks_d": 0.0453,
"share_hc_variance": 0.0168,
"rank_share_hc_variance": 7,
"rank_abs_studentized_resid": 5,
"rank_cooks_d": 7,
"dlog_gdp": 0.325,
"dw": 0.3758
}
},
"weighted": {
"seed": 20261250,
"full": {
"n": 27,
"b": 0.3078,
"se_hc1": 0.1414,
"p_hc1": 0.0295,
"se_hc3": 0.2194,
"p_hc3": 0.1606,
"p_wild": 0.2224
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"Japan",
"Taiwan",
"United Arab Emirates"
],
"n": 24,
"b": 0.3418,
"p_hc3": 0.2288,
"p_wild": 0.2128
},
"largest_share_hc_variance": {
"country": "Japan",
"value": 0.3883
},
"without_malta": {
"n": 26,
"b": 0.3076,
"se_hc1": 0.1414,
"p_hc1": 0.0296,
"se_hc3": 0.219,
"p_hc3": 0.1602,
"p_wild": 0.2219
},
"malta": {
"leverage": 0.0004,
"mean_leverage": 0.0741,
"studentized_resid": 0.2896,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 26,
"rank_abs_studentized_resid": 18,
"rank_cooks_d": 25,
"dlog_gdp": 0.325,
"dw": 0.3758,
"weight_share": 0.00043
}
}
},
"2007-2017": {
"unweighted": {
"seed": 20261259,
"full": {
"n": 27,
"b": 0.6256,
"se_hc1": 0.5391,
"p_hc1": 0.2459,
"se_hc3": 0.6564,
"p_hc3": 0.3406,
"p_wild": 0.3462
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"Malta",
"United Arab Emirates"
],
"n": 25,
"b": 0.3441,
"p_hc3": 0.0493,
"p_wild": 0.032
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.6401
},
"without_malta": {
"n": 26,
"b": 0.1435,
"se_hc1": 0.1708,
"p_hc1": 0.4007,
"se_hc3": 0.2314,
"p_hc3": 0.5351,
"p_wild": 0.4846
},
"malta": {
"leverage": 0.1309,
"mean_leverage": 0.0741,
"studentized_resid": 4.3377,
"cooks_d": 1.4173,
"share_hc_variance": 0.6401,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.4536,
"dw": 1.5007
}
},
"weighted": {
"seed": 20261260,
"full": {
"n": 27,
"b": -0.0648,
"se_hc1": 0.113,
"p_hc1": 0.5661,
"se_hc3": 0.1371,
"p_hc3": 0.6363,
"p_wild": 0.5773
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"South Korea"
],
"n": 26,
"b": 0.0099,
"p_hc3": 0.9575,
"p_wild": 0.9503
},
"largest_share_hc_variance": {
"country": "South Korea",
"value": 0.3099
},
"without_malta": {
"n": 26,
"b": -0.0773,
"se_hc1": 0.1101,
"p_hc1": 0.4828,
"se_hc3": 0.1323,
"p_hc3": 0.559,
"p_wild": 0.5153
},
"malta": {
"leverage": 0.003,
"mean_leverage": 0.0741,
"studentized_resid": 1.7355,
"cooks_d": 0.0045,
"share_hc_variance": 0.013,
"rank_share_hc_variance": 9,
"rank_abs_studentized_resid": 2,
"rank_cooks_d": 17,
"dlog_gdp": 0.4536,
"dw": 1.5007,
"weight_share": 0.00043
}
}
},
"2017-2022": {
"unweighted": {
"seed": 20261269,
"full": {
"n": 27,
"b": -0.4263,
"se_hc1": 0.4184,
"p_hc1": 0.3082,
"se_hc3": 0.4903,
"p_hc3": 0.3846,
"p_wild": 0.5308
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"Hong Kong"
],
"n": 26,
"b": 0.0282,
"p_hc3": 0.8677,
"p_wild": 0.8571
},
"largest_share_hc_variance": {
"country": "Hong Kong",
"value": 0.8335
},
"without_malta": {
"n": 26,
"b": -0.4585,
"se_hc1": 0.4783,
"p_hc1": 0.3378,
"se_hc3": 0.5697,
"p_hc3": 0.4209,
"p_wild": 0.5892
},
"malta": {
"leverage": 0.1898,
"mean_leverage": 0.0741,
"studentized_resid": 0.2661,
"cooks_d": 0.0083,
"share_hc_variance": 0.0041,
"rank_share_hc_variance": 10,
"rank_abs_studentized_resid": 21,
"rank_cooks_d": 14,
"dlog_gdp": 0.2156,
"dw": 0.0747
}
},
"weighted": {
"seed": 20261270,
"full": {
"n": 27,
"b": -0.0468,
"se_hc1": 0.1622,
"p_hc1": 0.773,
"se_hc3": 0.2014,
"p_hc3": 0.8163,
"p_wild": 0.7751
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1481,
"dropped": [
"Canada",
"Japan"
],
"n": 25,
"b": -0.1321,
"p_hc3": 0.5586,
"p_wild": 0.524
},
"largest_share_hc_variance": {
"country": "Hong Kong",
"value": 0.2708
},
"without_malta": {
"n": 26,
"b": -0.0468,
"se_hc1": 0.1629,
"p_hc1": 0.7738,
"se_hc3": 0.202,
"p_hc3": 0.8167,
"p_wild": 0.7737
},
"malta": {
"leverage": 0.0027,
"mean_leverage": 0.0741,
"studentized_resid": 0.0029,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 27,
"rank_abs_studentized_resid": 27,
"rank_cooks_d": 27,
"dlog_gdp": 0.2156,
"dw": 0.0747,
"weight_share": 0.00026
}
}
}
},
"core_no_norway": {
"1990-2007": {
"unweighted": {
"seed": 20261129,
"full": {
"n": 28,
"b": 0.3055,
"se_hc1": 0.2337,
"p_hc1": 0.1912,
"se_hc3": 0.2946,
"p_hc3": 0.2998,
"p_wild": 0.3367
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Belgium",
"Singapore"
],
"n": 26,
"b": 0.145,
"p_hc3": 0.3962,
"p_wild": 0.4039
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.672
},
"without_malta": {
"n": 27,
"b": 0.2974,
"se_hc1": 0.2382,
"p_hc1": 0.2119,
"se_hc3": 0.3025,
"p_hc3": 0.3256,
"p_wild": 0.3639
},
"malta": {
"leverage": 0.0382,
"mean_leverage": 0.0714,
"studentized_resid": 0.8714,
"cooks_d": 0.0151,
"share_hc_variance": 0.0012,
"rank_share_hc_variance": 17,
"rank_abs_studentized_resid": 11,
"rank_cooks_d": 14,
"dlog_gdp": 0.6769,
"dw": 0.3356
}
},
"weighted": {
"seed": 20261130,
"full": {
"n": 28,
"b": 0.2034,
"se_hc1": 0.168,
"p_hc1": 0.226,
"se_hc3": 0.2097,
"p_hc3": 0.3323,
"p_wild": 0.3613
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Japan",
"Singapore",
"South Korea"
],
"n": 25,
"b": 0.0593,
"p_hc3": 0.8522,
"p_wild": 0.8747
},
"largest_share_hc_variance": {
"country": "South Korea",
"value": 0.3627
},
"without_malta": {
"n": 27,
"b": 0.2029,
"se_hc1": 0.1683,
"p_hc1": 0.2279,
"se_hc3": 0.2099,
"p_hc3": 0.3337,
"p_wild": 0.3477
},
"malta": {
"leverage": 0.0006,
"mean_leverage": 0.0714,
"studentized_resid": 0.1566,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 23,
"rank_abs_studentized_resid": 24,
"rank_cooks_d": 27,
"dlog_gdp": 0.6769,
"dw": 0.3356,
"weight_share": 0.00022
}
}
},
"2007-2022": {
"unweighted": {
"seed": 20261139,
"full": {
"n": 28,
"b": 0.7809,
"se_hc1": 0.5423,
"p_hc1": 0.1499,
"se_hc3": 0.6698,
"p_hc3": 0.2437,
"p_wild": 0.1213
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Malta"
],
"n": 27,
"b": 0.1576,
"p_hc3": 0.2451,
"p_wild": 0.2326
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.8487
},
"without_malta": {
"n": 27,
"b": 0.1576,
"se_hc1": 0.1218,
"p_hc1": 0.1956,
"se_hc3": 0.1356,
"p_hc3": 0.2451,
"p_wild": 0.2326
},
"malta": {
"leverage": 0.2213,
"mean_leverage": 0.0714,
"studentized_resid": 3.9774,
"cooks_d": 2.2473,
"share_hc_variance": 0.8487,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.6692,
"dw": 1.5753
}
},
"weighted": {
"seed": 20261140,
"full": {
"n": 28,
"b": -0.0377,
"se_hc1": 0.082,
"p_hc1": 0.6454,
"se_hc3": 0.0943,
"p_hc3": 0.6891,
"p_wild": 0.6449
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [],
"n": 28,
"b": -0.0377,
"p_hc3": 0.6891,
"p_wild": 0.6449
},
"largest_share_hc_variance": {
"country": "South Korea",
"value": 0.2304
},
"without_malta": {
"n": 27,
"b": -0.05,
"se_hc1": 0.0818,
"p_hc1": 0.541,
"se_hc3": 0.0948,
"p_hc3": 0.598,
"p_wild": 0.5489
},
"malta": {
"leverage": 0.0038,
"mean_leverage": 0.0714,
"studentized_resid": 1.4304,
"cooks_d": 0.0039,
"share_hc_variance": 0.0238,
"rank_share_hc_variance": 10,
"rank_abs_studentized_resid": 5,
"rank_cooks_d": 16,
"dlog_gdp": 0.6692,
"dw": 1.5753,
"weight_share": 0.00021
}
}
},
"1997-2007": {
"unweighted": {
"seed": 20261149,
"full": {
"n": 28,
"b": 0.325,
"se_hc1": 0.3423,
"p_hc1": 0.3424,
"se_hc3": 0.378,
"p_hc3": 0.3898,
"p_wild": 0.3646
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Belgium",
"Singapore"
],
"n": 26,
"b": 0.1738,
"p_hc3": 0.537,
"p_wild": 0.5076
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.5866
},
"without_malta": {
"n": 27,
"b": 0.3481,
"se_hc1": 0.344,
"p_hc1": 0.3115,
"se_hc3": 0.3795,
"p_hc3": 0.359,
"p_wild": 0.3307
},
"malta": {
"leverage": 0.0391,
"mean_leverage": 0.0714,
"studentized_resid": 1.2037,
"cooks_d": 0.0295,
"share_hc_variance": 0.0045,
"rank_share_hc_variance": 10,
"rank_abs_studentized_resid": 6,
"rank_cooks_d": 7,
"dlog_gdp": 0.325,
"dw": 0.3758
}
},
"weighted": {
"seed": 20261150,
"full": {
"n": 28,
"b": 0.3771,
"se_hc1": 0.2248,
"p_hc1": 0.0934,
"se_hc3": 0.403,
"p_hc3": 0.3495,
"p_wild": 0.468
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Japan",
"Taiwan"
],
"n": 26,
"b": 0.2221,
"p_hc3": 0.462,
"p_wild": 0.5274
},
"largest_share_hc_variance": {
"country": "Japan",
"value": 0.4717
},
"without_malta": {
"n": 27,
"b": 0.3768,
"se_hc1": 0.2251,
"p_hc1": 0.0941,
"se_hc3": 0.403,
"p_hc3": 0.3497,
"p_wild": 0.4629
},
"malta": {
"leverage": 0.0002,
"mean_leverage": 0.0714,
"studentized_resid": 0.2244,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 27,
"rank_abs_studentized_resid": 20,
"rank_cooks_d": 27,
"dlog_gdp": 0.325,
"dw": 0.3758,
"weight_share": 0.00021
}
}
},
"2007-2017": {
"unweighted": {
"seed": 20261159,
"full": {
"n": 28,
"b": 1.0243,
"se_hc1": 0.6708,
"p_hc1": 0.1268,
"se_hc3": 0.8158,
"p_hc3": 0.2092,
"p_wild": 0.0514
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Greece",
"Malta",
"Singapore"
],
"n": 25,
"b": 0.5231,
"p_hc3": 0.2276,
"p_wild": 0.2117
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.7438
},
"without_malta": {
"n": 27,
"b": 0.3387,
"se_hc1": 0.1674,
"p_hc1": 0.0431,
"se_hc3": 0.1839,
"p_hc3": 0.0656,
"p_wild": 0.0436
},
"malta": {
"leverage": 0.1802,
"mean_leverage": 0.0714,
"studentized_resid": 4.3702,
"cooks_d": 2.0989,
"share_hc_variance": 0.7438,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.4536,
"dw": 1.5007
}
},
"weighted": {
"seed": 20261160,
"full": {
"n": 28,
"b": 0.0622,
"se_hc1": 0.165,
"p_hc1": 0.7062,
"se_hc3": 0.1902,
"p_hc3": 0.7436,
"p_wild": 0.7463
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"South Korea",
"United States"
],
"n": 26,
"b": 0.1896,
"p_hc3": 0.1806,
"p_wild": 0.1564
},
"largest_share_hc_variance": {
"country": "South Korea",
"value": 0.5858
},
"without_malta": {
"n": 27,
"b": 0.046,
"se_hc1": 0.1635,
"p_hc1": 0.7786,
"se_hc3": 0.1884,
"p_hc3": 0.8073,
"p_wild": 0.8102
},
"malta": {
"leverage": 0.0037,
"mean_leverage": 0.0714,
"studentized_resid": 1.6275,
"cooks_d": 0.0049,
"share_hc_variance": 0.0104,
"rank_share_hc_variance": 10,
"rank_abs_studentized_resid": 3,
"rank_cooks_d": 16,
"dlog_gdp": 0.4536,
"dw": 1.5007,
"weight_share": 0.00021
}
}
},
"2017-2022": {
"unweighted": {
"seed": 20261169,
"full": {
"n": 28,
"b": -0.609,
"se_hc1": 0.405,
"p_hc1": 0.1326,
"se_hc3": 0.474,
"p_hc3": 0.1988,
"p_wild": 0.1571
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Hong Kong"
],
"n": 27,
"b": -0.1784,
"p_hc3": 0.2787,
"p_wild": 0.2621
},
"largest_share_hc_variance": {
"country": "Hong Kong",
"value": 0.8035
},
"without_malta": {
"n": 27,
"b": -0.6883,
"se_hc1": 0.4563,
"p_hc1": 0.1314,
"se_hc3": 0.5429,
"p_hc3": 0.2049,
"p_wild": 0.1956
},
"malta": {
"leverage": 0.1931,
"mean_leverage": 0.0714,
"studentized_resid": 0.6708,
"cooks_d": 0.0539,
"share_hc_variance": 0.0265,
"rank_share_hc_variance": 4,
"rank_abs_studentized_resid": 12,
"rank_cooks_d": 4,
"dlog_gdp": 0.2156,
"dw": 0.0747
}
},
"weighted": {
"seed": 20261170,
"full": {
"n": 28,
"b": -0.3423,
"se_hc1": 0.1745,
"p_hc1": 0.0498,
"se_hc3": 0.2838,
"p_hc3": 0.2278,
"p_wild": 0.1676
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Japan",
"United States"
],
"n": 26,
"b": -0.3011,
"p_hc3": 0.2575,
"p_wild": 0.1632
},
"largest_share_hc_variance": {
"country": "United States",
"value": 0.225
},
"without_malta": {
"n": 27,
"b": -0.3429,
"se_hc1": 0.175,
"p_hc1": 0.0501,
"se_hc3": 0.2844,
"p_hc3": 0.228,
"p_wild": 0.166
},
"malta": {
"leverage": 0.0014,
"mean_leverage": 0.0714,
"studentized_resid": 0.0859,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 25,
"rank_abs_studentized_resid": 27,
"rank_cooks_d": 27,
"dlog_gdp": 0.2156,
"dw": 0.0747,
"weight_share": 0.00014
}
}
}
},
"net_importers_1990": {
"1990-2007": {
"unweighted": {
"seed": 20270201,
"full": {
"n": 28,
"b": 0.1049,
"se_hc1": 0.1197,
"p_hc1": 0.3811,
"se_hc3": 0.1301,
"p_hc3": 0.4204,
"p_wild": 0.3221
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Latvia",
"Singapore"
],
"n": 26,
"b": 0.0898,
"p_hc3": 0.4122,
"p_wild": 0.4937
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.5108
},
"without_malta": {
"n": 27,
"b": 0.1024,
"se_hc1": 0.1213,
"p_hc1": 0.3988,
"se_hc3": 0.1323,
"p_hc3": 0.4393,
"p_wild": 0.3475
},
"malta": {
"leverage": 0.0363,
"mean_leverage": 0.0714,
"studentized_resid": 0.8981,
"cooks_d": 0.0152,
"share_hc_variance": 0.0004,
"rank_share_hc_variance": 17,
"rank_abs_studentized_resid": 9,
"rank_cooks_d": 9,
"dlog_gdp": 0.6769,
"dw": 0.3356
}
},
"weighted": {
"seed": 20270202,
"full": {
"n": 28,
"b": 0.1222,
"se_hc1": 0.0982,
"p_hc1": 0.2131,
"se_hc3": 0.1186,
"p_hc3": 0.3026,
"p_wild": 0.2925
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Japan",
"Singapore",
"South Korea",
"United Kingdom"
],
"n": 24,
"b": 0.0683,
"p_hc3": 0.7437,
"p_wild": 0.7636
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.4167
},
"without_malta": {
"n": 27,
"b": 0.1219,
"se_hc1": 0.0983,
"p_hc1": 0.2153,
"se_hc3": 0.1186,
"p_hc3": 0.3042,
"p_wild": 0.2922
},
"malta": {
"leverage": 0.0008,
"mean_leverage": 0.0714,
"studentized_resid": 0.2204,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 24,
"rank_abs_studentized_resid": 23,
"rank_cooks_d": 28,
"dlog_gdp": 0.6769,
"dw": 0.3356,
"weight_share": 0.00044
}
}
},
"1997-2007": {
"unweighted": {
"seed": 20270211,
"full": {
"n": 28,
"b": 0.2935,
"se_hc1": 0.1595,
"p_hc1": 0.0658,
"se_hc3": 0.2906,
"p_hc3": 0.3125,
"p_wild": 0.0971
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Singapore",
"United Arab Emirates"
],
"n": 26,
"b": 0.384,
"p_hc3": 0.0343,
"p_wild": 0.0828
},
"largest_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.5426
},
"without_malta": {
"n": 27,
"b": 0.3127,
"se_hc1": 0.1625,
"p_hc1": 0.0543,
"se_hc3": 0.3006,
"p_hc3": 0.2982,
"p_wild": 0.1035
},
"malta": {
"leverage": 0.0424,
"mean_leverage": 0.0714,
"studentized_resid": 1.4714,
"cooks_d": 0.048,
"share_hc_variance": 0.0141,
"rank_share_hc_variance": 7,
"rank_abs_studentized_resid": 5,
"rank_cooks_d": 5,
"dlog_gdp": 0.325,
"dw": 0.3758
}
},
"weighted": {
"seed": 20270212,
"full": {
"n": 28,
"b": 0.3772,
"se_hc1": 0.1462,
"p_hc1": 0.0099,
"se_hc3": 0.2478,
"p_hc3": 0.128,
"p_wild": 0.1658
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1429,
"dropped": [
"Japan",
"Singapore",
"United Arab Emirates"
],
"n": 25,
"b": 0.108,
"p_hc3": 0.457,
"p_wild": 0.4166
},
"largest_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.4556
},
"without_malta": {
"n": 27,
"b": 0.3771,
"se_hc1": 0.1462,
"p_hc1": 0.0099,
"se_hc3": 0.2475,
"p_hc3": 0.1276,
"p_wild": 0.1709
},
"malta": {
"leverage": 0.0005,
"mean_leverage": 0.0714,
"studentized_resid": 0.3332,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 27,
"rank_abs_studentized_resid": 20,
"rank_cooks_d": 27,
"dlog_gdp": 0.325,
"dw": 0.3758,
"weight_share": 0.00044
}
}
}
}
},
"table6": {
"avg3_1990-92_to_2005-07": {
"unweighted": {
"seed": 20270301,
"full": {
"n": 37,
"b": 0.0464,
"se_hc1": 0.2491,
"p_hc1": 0.8522,
"se_hc3": 0.282,
"p_hc3": 0.8693,
"p_wild": 0.8823
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1081,
"dropped": [
"Belgium",
"Latvia",
"Singapore",
"Trinidad and Tobago"
],
"n": 33,
"b": 0.0639,
"p_hc3": 0.7162,
"p_wild": 0.6749
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.5729
},
"without_malta": {
"n": 36,
"b": 0.045,
"se_hc1": 0.2504,
"p_hc1": 0.8573,
"se_hc3": 0.284,
"p_hc3": 0.874,
"p_wild": 0.8929
},
"malta": {
"leverage": 0.0286,
"mean_leverage": 0.0541,
"studentized_resid": 0.1907,
"cooks_d": 0.0005,
"share_hc_variance": 0,
"rank_share_hc_variance": 30,
"rank_abs_studentized_resid": 30,
"rank_cooks_d": 30,
"dlog_gdp": 0.5941,
"dw": 0.1306
}
},
"weighted": {
"seed": 20270302,
"full": {
"n": 37,
"b": 0.093,
"se_hc1": 0.1584,
"p_hc1": 0.557,
"se_hc3": 0.2135,
"p_hc3": 0.6629,
"p_wild": 0.618
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1081,
"dropped": [
"Japan",
"Saudi Arabia",
"Singapore",
"South Korea",
"Taiwan",
"United States"
],
"n": 31,
"b": -0.1931,
"p_hc3": 0.4615,
"p_wild": 0.428
},
"largest_share_hc_variance": {
"country": "Saudi Arabia",
"value": 0.3343
},
"without_malta": {
"n": 36,
"b": 0.093,
"se_hc1": 0.1586,
"p_hc1": 0.5576,
"se_hc3": 0.2135,
"p_hc3": 0.6631,
"p_wild": 0.6267
},
"malta": {
"leverage": 0.0003,
"mean_leverage": 0.0541,
"studentized_resid": 0.0405,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 37,
"rank_abs_studentized_resid": 36,
"rank_cooks_d": 37,
"dlog_gdp": 0.5941,
"dw": 0.1306,
"weight_share": 0.0002
}
}
},
"benchmark_1997-2007": {
"unweighted": {
"seed": 20261049,
"full": {
"n": 39,
"b": 0.0574,
"se_hc1": 0.12,
"p_hc1": 0.6325,
"se_hc3": 0.1256,
"p_hc3": 0.6479,
"p_wild": 0.6125
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1026,
"dropped": [
"Belgium",
"Singapore",
"Trinidad and Tobago"
],
"n": 36,
"b": 0.129,
"p_hc3": 0.1523,
"p_wild": 0.1526
},
"largest_share_hc_variance": {
"country": "Trinidad and Tobago",
"value": 0.2424
},
"without_malta": {
"n": 38,
"b": 0.074,
"se_hc1": 0.1194,
"p_hc1": 0.5357,
"se_hc3": 0.1248,
"p_hc3": 0.5535,
"p_wild": 0.5159
},
"malta": {
"leverage": 0.0333,
"mean_leverage": 0.0513,
"studentized_resid": 1.262,
"cooks_d": 0.0274,
"share_hc_variance": 0.0188,
"rank_share_hc_variance": 11,
"rank_abs_studentized_resid": 7,
"rank_cooks_d": 10,
"dlog_gdp": 0.325,
"dw": 0.3758
}
},
"weighted": {
"seed": 20261050,
"full": {
"n": 39,
"b": 0.1369,
"se_hc1": 0.1387,
"p_hc1": 0.3235,
"se_hc3": 0.182,
"p_hc3": 0.4519,
"p_wild": 0.5036
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1026,
"dropped": [
"Japan",
"Saudi Arabia"
],
"n": 37,
"b": 0.0698,
"p_hc3": 0.5249,
"p_wild": 0.4641
},
"largest_share_hc_variance": {
"country": "Japan",
"value": 0.5587
},
"without_malta": {
"n": 38,
"b": 0.1369,
"se_hc1": 0.1387,
"p_hc1": 0.3237,
"se_hc3": 0.182,
"p_hc3": 0.4518,
"p_wild": 0.4994
},
"malta": {
"leverage": 0.0002,
"mean_leverage": 0.0513,
"studentized_resid": 0.2582,
"cooks_d": 0,
"share_hc_variance": 0,
"rank_share_hc_variance": 39,
"rank_abs_studentized_resid": 25,
"rank_cooks_d": 38,
"dlog_gdp": 0.325,
"dw": 0.3758,
"weight_share": 0.0002
}
}
},
"avg3_2005-07_to_2020-22": {
"unweighted": {
"seed": 20270321,
"full": {
"n": 40,
"b": 0.0962,
"se_hc1": 0.2093,
"p_hc1": 0.6458,
"se_hc3": 0.2583,
"p_hc3": 0.7096,
"p_wild": 0.876
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Malta",
"United Arab Emirates",
"Venezuela"
],
"n": 37,
"b": 0.0012,
"p_hc3": 0.992,
"p_wild": 0.9916
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.7319
},
"without_malta": {
"n": 39,
"b": -0.0906,
"se_hc1": 0.0573,
"p_hc1": 0.1138,
"se_hc3": 0.0616,
"p_hc3": 0.1414,
"p_wild": 0.1758
},
"malta": {
"leverage": 0.0658,
"mean_leverage": 0.05,
"studentized_resid": 5.2463,
"cooks_d": 0.97,
"share_hc_variance": 0.7319,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.6205,
"dw": 1.764
}
},
"weighted": {
"seed": 20270322,
"full": {
"n": 40,
"b": -0.0756,
"se_hc1": 0.0572,
"p_hc1": 0.1862,
"se_hc3": 0.0714,
"p_hc3": 0.2894,
"p_wild": 0.3544
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Japan",
"Saudi Arabia",
"United States",
"Venezuela"
],
"n": 36,
"b": -0.1941,
"p_hc3": 0.0171,
"p_wild": 0.05
},
"largest_share_hc_variance": {
"country": "Saudi Arabia",
"value": 0.5197
},
"without_malta": {
"n": 39,
"b": -0.0789,
"se_hc1": 0.0572,
"p_hc1": 0.1676,
"se_hc3": 0.0709,
"p_hc3": 0.2658,
"p_wild": 0.3338
},
"malta": {
"leverage": 0.0009,
"mean_leverage": 0.05,
"studentized_resid": 1.87,
"cooks_d": 0.0016,
"share_hc_variance": 0.0034,
"rank_share_hc_variance": 14,
"rank_abs_studentized_resid": 3,
"rank_cooks_d": 23,
"dlog_gdp": 0.6205,
"dw": 1.764,
"weight_share": 0.00019
}
}
},
"benchmark_2007-2017": {
"unweighted": {
"seed": 20261059,
"full": {
"n": 40,
"b": 0.437,
"se_hc1": 0.3443,
"p_hc1": 0.2042,
"se_hc3": 0.3788,
"p_hc3": 0.2486,
"p_wild": 0.2127
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Malta",
"United Arab Emirates"
],
"n": 38,
"b": 0.2275,
"p_hc3": 0.0334,
"p_wild": 0.0371
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.7223
},
"without_malta": {
"n": 39,
"b": 0.1247,
"se_hc1": 0.1222,
"p_hc1": 0.3077,
"se_hc3": 0.1399,
"p_hc3": 0.3728,
"p_wild": 0.348
},
"malta": {
"leverage": 0.0873,
"mean_leverage": 0.05,
"studentized_resid": 5.1505,
"cooks_d": 1.2686,
"share_hc_variance": 0.7223,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.4536,
"dw": 1.5007
}
},
"weighted": {
"seed": 20261060,
"full": {
"n": 40,
"b": 0.0804,
"se_hc1": 0.0902,
"p_hc1": 0.373,
"se_hc3": 0.1104,
"p_hc3": 0.4665,
"p_wild": 0.4562
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Saudi Arabia",
"South Korea",
"United Arab Emirates",
"United States"
],
"n": 36,
"b": 0.1922,
"p_hc3": 0.0526,
"p_wild": 0.0951
},
"largest_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.325
},
"without_malta": {
"n": 39,
"b": 0.0752,
"se_hc1": 0.0908,
"p_hc1": 0.4075,
"se_hc3": 0.112,
"p_hc3": 0.502,
"p_wild": 0.4849
},
"malta": {
"leverage": 0.0013,
"mean_leverage": 0.05,
"studentized_resid": 1.7501,
"cooks_d": 0.0019,
"share_hc_variance": 0.0034,
"rank_share_hc_variance": 12,
"rank_abs_studentized_resid": 3,
"rank_cooks_d": 22,
"dlog_gdp": 0.4536,
"dw": 1.5007,
"weight_share": 0.00019
}
}
},
"end2019_2007-2019": {
"unweighted": {
"seed": 20270341,
"full": {
"n": 40,
"b": 0.5355,
"se_hc1": 0.2745,
"p_hc1": 0.0511,
"se_hc3": 0.3069,
"p_hc3": 0.081,
"p_wild": 0.2719
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Malta",
"United Arab Emirates",
"Venezuela"
],
"n": 37,
"b": 0.0707,
"p_hc3": 0.6275,
"p_wild": 0.6271
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.7242
},
"without_malta": {
"n": 39,
"b": 0.2856,
"se_hc1": 0.1912,
"p_hc1": 0.1352,
"se_hc3": 0.3407,
"p_hc3": 0.4019,
"p_wild": 0.592
},
"malta": {
"leverage": 0.089,
"mean_leverage": 0.05,
"studentized_resid": 5.047,
"cooks_d": 1.2443,
"share_hc_variance": 0.7242,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 1,
"dlog_gdp": 0.5814,
"dw": 1.6327
}
},
"weighted": {
"seed": 20270342,
"full": {
"n": 40,
"b": 0.238,
"se_hc1": 0.1747,
"p_hc1": 0.1729,
"se_hc3": 0.282,
"p_hc3": 0.3986,
"p_wild": 0.6013
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Italy",
"United Arab Emirates",
"United States",
"Venezuela"
],
"n": 36,
"b": 0.066,
"p_hc3": 0.43,
"p_wild": 0.4568
},
"largest_share_hc_variance": {
"country": "Venezuela",
"value": 0.8248
},
"without_malta": {
"n": 39,
"b": 0.2338,
"se_hc1": 0.1761,
"p_hc1": 0.1844,
"se_hc3": 0.285,
"p_hc3": 0.412,
"p_wild": 0.6316
},
"malta": {
"leverage": 0.0013,
"mean_leverage": 0.05,
"studentized_resid": 1.5287,
"cooks_d": 0.0015,
"share_hc_variance": 0.0006,
"rank_share_hc_variance": 12,
"rank_abs_studentized_resid": 6,
"rank_cooks_d": 20,
"dlog_gdp": 0.5814,
"dw": 1.6327,
"weight_share": 0.00019
}
}
}
},
"table5": {
"start1995_pre": {
"seed": 20270401,
"full": {
"n": 39,
"b": 0.1009,
"se_hc1": 0.0957,
"p_hc1": 0.2915,
"se_hc3": 0.1036,
"p_hc3": 0.3302,
"p_wild": 0.3243
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1026,
"dropped": [
"Belgium",
"Trinidad and Tobago"
],
"n": 37,
"b": 0.1855,
"p_hc3": 0.0306,
"p_wild": 0.03
},
"largest_share_hc_variance": {
"country": "Belgium",
"value": 0.225
},
"without_malta": {
"n": 38,
"b": 0.1059,
"se_hc1": 0.096,
"p_hc1": 0.2701,
"se_hc3": 0.1038,
"p_hc3": 0.3077,
"p_wild": 0.3026
},
"malta": {
"leverage": 0.0315,
"mean_leverage": 0.0513,
"studentized_resid": 0.5566,
"cooks_d": 0.005,
"share_hc_variance": 0.0027,
"rank_share_hc_variance": 21,
"rank_abs_studentized_resid": 21,
"rank_cooks_d": 24,
"dlog_gdp": 0.4131,
"dw": 0.1927
}
},
"pop1m_pre": {
"seed": 20270411,
"full": {
"n": 35,
"b": 0.0343,
"se_hc1": 0.1175,
"p_hc1": 0.7705,
"se_hc3": 0.1365,
"p_hc3": 0.8017,
"p_wild": 0.7665
},
"malta_in_sample": false,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1143,
"dropped": [
"Latvia",
"Singapore",
"United Arab Emirates"
],
"n": 32,
"b": -0.1294,
"p_hc3": 0.4552,
"p_wild": 0.4086
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.4249
}
},
"pop1m_post": {
"seed": 20270421,
"full": {
"n": 37,
"b": -0.0087,
"se_hc1": 0.0535,
"p_hc1": 0.8711,
"se_hc3": 0.0574,
"p_hc3": 0.8798,
"p_wild": 0.859
},
"malta_in_sample": false,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1081,
"dropped": [
"Hong Kong",
"Norway"
],
"n": 35,
"b": 0.0091,
"p_hc3": 0.8759,
"p_wild": 0.854
},
"largest_share_hc_variance": {
"country": "United Arab Emirates",
"value": 0.4269
}
},
"percapita_pre": {
"seed": 20270431,
"full": {
"n": 38,
"b": 0.1456,
"se_hc1": 0.2095,
"p_hc1": 0.487,
"se_hc3": 0.2229,
"p_hc3": 0.5135,
"p_wild": 0.4949
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1053,
"dropped": [
"Belgium",
"Latvia",
"Singapore",
"Trinidad and Tobago"
],
"n": 34,
"b": 0.2352,
"p_hc3": 0.0588,
"p_wild": 0.1538
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.4018
},
"without_malta": {
"n": 37,
"b": 0.1373,
"se_hc1": 0.2119,
"p_hc1": 0.5172,
"se_hc3": 0.2264,
"p_hc3": 0.5444,
"p_wild": 0.5238
},
"malta": {
"leverage": 0.0288,
"mean_leverage": 0.0526,
"studentized_resid": 0.8565,
"cooks_d": 0.0109,
"share_hc_variance": 0.0016,
"rank_share_hc_variance": 19,
"rank_abs_studentized_resid": 12,
"rank_cooks_d": 14,
"dlog_gdp": 0.5511,
"dw": 0.3356
}
},
"percapita_post": {
"seed": 20270441,
"full": {
"n": 40,
"b": 0.208,
"se_hc1": 0.2087,
"p_hc1": 0.3188,
"se_hc3": 0.3284,
"p_hc3": 0.5264,
"p_wild": 0.4211
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1,
"dropped": [
"Malta",
"Venezuela"
],
"n": 38,
"b": 0.0685,
"p_hc3": 0.3887,
"p_wild": 0.3755
},
"largest_share_hc_variance": {
"country": "Malta",
"value": 0.7057
},
"without_malta": {
"n": 39,
"b": 0.0254,
"se_hc1": 0.0396,
"p_hc1": 0.5214,
"se_hc3": 0.057,
"p_hc3": 0.6558,
"p_wild": 0.5419
},
"malta": {
"leverage": 0.0647,
"mean_leverage": 0.05,
"studentized_resid": 4.7813,
"cooks_d": 0.7913,
"share_hc_variance": 0.7057,
"rank_share_hc_variance": 1,
"rank_abs_studentized_resid": 1,
"rank_cooks_d": 2,
"dlog_gdp": 0.4333,
"dw": 1.5753
}
},
"kuwait_keep_all_pre": {
"seed": 20270451,
"full": {
"n": 39,
"b": 0.002,
"se_hc1": 0.0884,
"p_hc1": 0.9817,
"se_hc3": 0.1001,
"p_hc3": 0.9838,
"p_wild": 0.9804
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1026,
"dropped": [
"Latvia",
"Singapore",
"United Arab Emirates"
],
"n": 36,
"b": -0.0807,
"p_hc3": 0.357,
"p_wild": 0.1682
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.3383
},
"without_malta": {
"n": 38,
"b": 0.0021,
"se_hc1": 0.0892,
"p_hc1": 0.9814,
"se_hc3": 0.1012,
"p_hc3": 0.9836,
"p_wild": 0.9819
},
"malta": {
"leverage": 0.0256,
"mean_leverage": 0.0513,
"studentized_resid": 0.9067,
"cooks_d": 0.0108,
"share_hc_variance": 0,
"rank_share_hc_variance": 38,
"rank_abs_studentized_resid": 10,
"rank_cooks_d": 17,
"dlog_gdp": 0.6769,
"dw": 0.3356
}
},
"kuwait_drop1991_92_pre": {
"seed": 20270451,
"full": {
"n": 39,
"b": 0.002,
"se_hc1": 0.0884,
"p_hc1": 0.9817,
"se_hc3": 0.1001,
"p_hc3": 0.9838,
"p_wild": 0.9804
},
"malta_in_sample": true,
"screen_cooks_d_4_over_n": {
"cutoff": 0.1026,
"dropped": [
"Latvia",
"Singapore",
"United Arab Emirates"
],
"n": 36,
"b": -0.0807,
"p_hc3": 0.357,
"p_wild": 0.1682
},
"largest_share_hc_variance": {
"country": "Singapore",
"value": 0.3383
},
"without_malta": {
"n": 38,
"b": 0.0021,
"se_hc1": 0.0892,
"p_hc1": 0.9814,
"se_hc3": 0.1012,
"p_hc3": 0.9836,
"p_wild": 0.9819
},
"malta": {
"leverage": 0.0256,
"mean_leverage": 0.0513,
"studentized_resid": 0.9067,
"cooks_d": 0.0108,
"share_hc_variance": 0,
"rank_share_hc_variance": 38,
"rank_abs_studentized_resid": 10,
"rank_cooks_d": 17,
"dlog_gdp": 0.6769,
"dw": 0.3356
}
}
},
"table3_weighted_minus_unweighted_without_malta": {
"1990-2007": {
"diff_weighted_minus_unweighted": 0.0404,
"boot_se": 0.1294,
"boot_ci95_pct": [
-0.15,
0.3795
],
"dumouchel_duncan_gdp_x_w": {
"b": 0.0456,
"se": 0.3356,
"p": 0.8919,
"stars": "",
"se_type": "HC3"
},
"n": 37
},
"2007-2022": {
"diff_weighted_minus_unweighted": -0.0391,
"boot_se": 0.0627,
"boot_ci95_pct": [
-0.2013,
0.0647
],
"dumouchel_duncan_gdp_x_w": {
"b": -0.0605,
"se": 0.2127,
"p": 0.7762,
"stars": "",
"se_type": "HC3"
},
"n": 39
}
},
"period_comparison": {
"1990-2007_vs_2007-2022": {
"all": {
"full": {
"first": 0.0274,
"second": 0.1866,
"second_minus_first": 0.1592,
"second_exceeds_first": true,
"stacked": {
"change_high_wedge": 0.1592,
"se": 0.2475,
"p": 0.5201,
"n_obs": 235,
"n_countries": 119
},
"weighted_first": 0.0665,
"weighted_second": -0.0432,
"weighted_second_exceeds_first": false
},
"without_malta": {
"first": 0.0259,
"second": -0.0075,
"second_minus_first": -0.0334,
"second_exceeds_first": false,
"stacked": {
"change_high_wedge": -0.0333,
"se": 0.1286,
"p": 0.7954,
"n_obs": 233,
"n_countries": 118
},
"weighted_first": 0.0663,
"weighted_second": -0.0465,
"weighted_second_exceeds_first": false
}
},
"no_oil_transition": {
"full": {
"first": 0.3055,
"second": 0.7383,
"second_minus_first": 0.4328,
"second_exceeds_first": true,
"stacked": {
"change_high_wedge": 0.4328,
"se": 0.6184,
"p": 0.484,
"n_obs": 214,
"n_countries": 108
},
"weighted_first": 0.2034,
"weighted_second": -0.0522,
"weighted_second_exceeds_first": false
},
"without_malta": {
"first": 0.2974,
"second": 0.1018,
"second_minus_first": -0.1956,
"second_exceeds_first": false,
"stacked": {
"change_high_wedge": -0.1956,
"se": 0.2265,
"p": 0.3877,
"n_obs": 212,
"n_countries": 107
},
"weighted_first": 0.2029,
"weighted_second": -0.0645,
"weighted_second_exceeds_first": false
}
},
"core_no_norway": {
"full": {
"first": 0.3055,
"second": 0.7809,
"second_minus_first": 0.4754,
"second_exceeds_first": true,
"stacked": {
"change_high_wedge": 0.4754,
"se": 0.6167,
"p": 0.4407,
"n_obs": 213,
"n_countries": 107
},
"weighted_first": 0.2034,
"weighted_second": -0.0377,
"weighted_second_exceeds_first": false
},
"without_malta": {
"first": 0.2974,
"second": 0.1576,
"second_minus_first": -0.1398,
"second_exceeds_first": false,
"stacked": {
"change_high_wedge": -0.1398,
"se": 0.2341,
"p": 0.5506,
"n_obs": 211,
"n_countries": 106
},
"weighted_first": 0.2029,
"weighted_second": -0.05,
"weighted_second_exceeds_first": false
}
},
"net_importers_1997": {
"full": {
"first": 0.1303,
"second": 0.527,
"second_minus_first": 0.3967,
"second_exceeds_first": true,
"stacked": {
"change_high_wedge": 0.3967,
"se": 0.51,
"p": 0.4367,
"n_obs": 211,
"n_countries": 106
},
"weighted_first": 0.1716,
"weighted_second": -0.0894,
"weighted_second_exceeds_first": false
},
"without_malta": {
"first": 0.1282,
"second": 0.0208,
"second_minus_first": -0.1074,
"second_exceeds_first": false,
"stacked": {
"change_high_wedge": -0.1074,
"se": 0.1661,
"p": 0.5177,
"n_obs": 209,
"n_countries": 105
},
"weighted_first": 0.1713,
"weighted_second": -0.1016,
"weighted_second_exceeds_first": false
}
}
},
"1997-2007_vs_2007-2017": {
"all": {
"full": {
"first": 0.0574,
"second": 0.437,
"second_minus_first": 0.3796,
"second_exceeds_first": true,
"stacked": {
"change_high_wedge": 0.3797,
"se": 0.3806,
"p": 0.3185,
"n_obs": 237,
"n_countries": 119
},
"weighted_first": 0.1369,
"weighted_second": 0.0804,
"weighted_second_exceeds_first": false
},
"without_malta": {
"first": 0.074,
"second": 0.1247,
"second_minus_first": 0.0507,
"second_exceeds_first": true,
"stacked": {
"change_high_wedge": 0.0507,
"se": 0.1655,
"p": 0.7592,
"n_obs": 235,
"n_countries": 118
},
"weighted_first": 0.1369,
"weighted_second": 0.0752,
"weighted_second_exceeds_first": false
}
},
"no_oil_transition": {
"full": {
"first": 0.325,
"second": 0.9852,
"second_minus_first": 0.6602,
"second_exceeds_first": true,
"stacked": {
"change_high_wedge": 0.6602,
"se": 0.8319,
"p": 0.4274,
"n_obs": 215,
"n_countries": 108
},
"weighted_first": 0.3771,
"weighted_second": 0.0519,
"weighted_second_exceeds_first": false
},
"without_malta": {
"first": 0.3481,
"second": 0.3039,
"second_minus_first": -0.0442,
"second_exceeds_first": false,
"stacked": {
"change_high_wedge": -0.0442,
"se": 0.4086,
"p": 0.9138,
"n_obs": 213,
"n_countries": 107
},
"weighted_first": 0.3768,
"weighted_second": 0.0357,
"weighted_second_exceeds_first": false
}
},
"core_no_norway": {
"full": {
"first": 0.325,
"second": 1.0243,
"second_minus_first": 0.6993,
"second_exceeds_first": true,
"stacked": {
"change_high_wedge": 0.6993,
"se": 0.8439,
"p": 0.4073,
"n_obs": 214,
"n_countries": 107
},
"weighted_first": 0.3771,
"weighted_second": 0.0622,
"weighted_second_exceeds_first": false
},
"without_malta": {
"first": 0.3481,
"second": 0.3387,
"second_minus_first": -0.0094,
"second_exceeds_first": false,
"stacked": {
"change_high_wedge": -0.0095,
"se": 0.4185,
"p": 0.9819,
"n_obs": 212,
"n_countries": 106
},
"weighted_first": 0.3768,
"weighted_second": 0.046,
"weighted_second_exceeds_first": false
}
},
"net_importers_1997": {
"full": {
"first": 0.2786,
"second": 0.6256,
"second_minus_first": 0.347,
"second_exceeds_first": true,
"stacked": {
"change_high_wedge": 0.3469,
"se": 0.5394,
"p": 0.5201,
"n_obs": 212,
"n_countries": 106
},
"weighted_first": 0.3078,
"weighted_second": -0.0648,
"weighted_second_exceeds_first": false
},
"without_malta": {
"first": 0.2983,
"second": 0.1435,
"second_minus_first": -0.1548,
"second_exceeds_first": false,
"stacked": {
"change_high_wedge": -0.1548,
"se": 0.1662,
"p": 0.3516,
"n_obs": 210,
"n_countries": 105
},
"weighted_first": 0.3076,
"weighted_second": -0.0773,
"weighted_second_exceeds_first": false
}
}
}
},
"aggregate_wedge": {
"all": {
"1990-2007": {
"full": {
"n": 38,
"w_start": 0.0713,
"w_end": 0.1351,
"change": 0.0638
},
"without_malta": {
"n": 37,
"w_start": 0.0712,
"w_end": 0.1349,
"change": 0.0637
}
},
"2007-2022": {
"full": {
"n": 40,
"w_start": 0.1348,
"w_end": 0.1669,
"change": 0.0321
},
"without_malta": {
"n": 39,
"w_start": 0.1346,
"w_end": 0.1659,
"change": 0.0312
}
},
"1997-2007": {
"full": {
"n": 39,
"w_start": 0.0738,
"w_end": 0.1342,
"change": 0.0604
},
"without_malta": {
"n": 38,
"w_start": 0.0738,
"w_end": 0.1341,
"change": 0.0603
}
},
"2007-2017": {
"full": {
"n": 40,
"w_start": 0.1348,
"w_end": 0.1139,
"change": -0.0209
},
"without_malta": {
"n": 39,
"w_start": 0.1346,
"w_end": 0.1131,
"change": -0.0216
}
},
"2017-2022": {
"full": {
"n": 40,
"w_start": 0.1139,
"w_end": 0.1669,
"change": 0.053
},
"without_malta": {
"n": 39,
"w_start": 0.1131,
"w_end": 0.1659,
"change": 0.0528
}
},
"2007-2019": {
"full": {
"n": 40,
"w_start": 0.1348,
"w_end": 0.1231,
"change": -0.0117
},
"without_malta": {
"n": 39,
"w_start": 0.1346,
"w_end": 0.1221,
"change": -0.0125
}
}
},
"no_oil_transition": {
"1990-2007": {
"full": {
"n": 28,
"w_start": 0.0792,
"w_end": 0.1458,
"change": 0.0666
},
"without_malta": {
"n": 27,
"w_start": 0.0792,
"w_end": 0.1457,
"change": 0.0665
}
},
"2007-2022": {
"full": {
"n": 29,
"w_start": 0.1464,
"w_end": 0.1845,
"change": 0.0381
},
"without_malta": {
"n": 28,
"w_start": 0.1463,
"w_end": 0.1834,
"change": 0.0371
}
},
"1997-2007": {
"full": {
"n": 28,
"w_start": 0.0818,
"w_end": 0.1458,
"change": 0.064
},
"without_malta": {
"n": 27,
"w_start": 0.0818,
"w_end": 0.1457,
"change": 0.0639
}
},
"2007-2017": {
"full": {
"n": 29,
"w_start": 0.1464,
"w_end": 0.1273,
"change": -0.0191
},
"without_malta": {
"n": 28,
"w_start": 0.1463,
"w_end": 0.1264,
"change": -0.0199
}
},
"2017-2022": {
"full": {
"n": 29,
"w_start": 0.1273,
"w_end": 0.1845,
"change": 0.0572
},
"without_malta": {
"n": 28,
"w_start": 0.1264,
"w_end": 0.1834,
"change": 0.0569
}
}
},
"net_importers_1997": {
"1990-2007": {
"full": {
"n": 27,
"w_start": 0.1707,
"w_end": 0.2246,
"change": 0.0539
},
"without_malta": {
"n": 26,
"w_start": 0.1707,
"w_end": 0.2244,
"change": 0.0538
}
},
"2007-2022": {
"full": {
"n": 27,
"w_start": 0.2246,
"w_end": 0.2686,
"change": 0.044
},
"without_malta": {
"n": 26,
"w_start": 0.2244,
"w_end": 0.2665,
"change": 0.0421
}
},
"1997-2007": {
"full": {
"n": 27,
"w_start": 0.173,
"w_end": 0.2246,
"change": 0.0516
},
"without_malta": {
"n": 26,
"w_start": 0.1729,
"w_end": 0.2244,
"change": 0.0515
}
},
"2007-2017": {
"full": {
"n": 27,
"w_start": 0.2246,
"w_end": 0.1848,
"change": -0.0398
},
"without_malta": {
"n": 26,
"w_start": 0.2244,
"w_end": 0.1832,
"change": -0.0412
}
},
"2017-2022": {
"full": {
"n": 27,
"w_start": 0.1848,
"w_end": 0.2686,
"change": 0.0838
},
"without_malta": {
"n": 26,
"w_start": 0.1832,
"w_end": 0.2665,
"change": 0.0833
}
}
},
"core_no_norway": {
"1990-2007": {
"full": {
"n": 28,
"w_start": 0.0792,
"w_end": 0.1458,
"change": 0.0666
},
"without_malta": {
"n": 27,
"w_start": 0.0792,
"w_end": 0.1457,
"change": 0.0665
}
},
"2007-2022": {
"full": {
"n": 28,
"w_start": 0.1458,
"w_end": 0.1856,
"change": 0.0398
},
"without_malta": {
"n": 27,
"w_start": 0.1457,
"w_end": 0.1845,
"change": 0.0388
}
},
"1997-2007": {
"full": {
"n": 28,
"w_start": 0.0818,
"w_end": 0.1458,
"change": 0.064
},
"without_malta": {
"n": 27,
"w_start": 0.0818,
"w_end": 0.1457,
"change": 0.0639
}
},
"2007-2017": {
"full": {
"n": 28,
"w_start": 0.1458,
"w_end": 0.1274,
"change": -0.0185
},
"without_malta": {
"n": 27,
"w_start": 0.1457,
"w_end": 0.1265,
"change": -0.0192
}
},
"2017-2022": {
"full": {
"n": 28,
"w_start": 0.1274,
"w_end": 0.1856,
"change": 0.0582
},
"without_malta": {
"n": 27,
"w_start": 0.1265,
"w_end": 0.1845,
"change": 0.058
}
}
},
"net_importers_1990": {
"1990-2007": {
"full": {
"n": 28,
"w_start": 0.1804,
"w_end": 0.2325,
"change": 0.0521
},
"without_malta": {
"n": 27,
"w_start": 0.1804,
"w_end": 0.2323,
"change": 0.052
}
},
"1997-2007": {
"full": {
"n": 28,
"w_start": 0.169,
"w_end": 0.2325,
"change": 0.0635
},
"without_malta": {
"n": 27,
"w_start": 0.169,
"w_end": 0.2323,
"change": 0.0634
}
}
}
},
"summary": {
"n_rows": 61,
"n_distinct_rows": 52,
"n_rows_with_malta": 59,
"rows_without_malta_in_sample": [
"table5|pop1m_pre|unweighted",
"table5|pop1m_post|unweighted"
],
"distinct_malta_largest_share_hc_variance": [
"table4|all|2007-2017|unweighted",
"table4|all|2007-2022|unweighted",
"table4|core_no_norway|2007-2017|unweighted",
"table4|core_no_norway|2007-2022|unweighted",
"table4|net_importers_1997|2007-2017|unweighted",
"table4|net_importers_1997|2007-2022|unweighted",
"table4|no_oil_transition|2007-2017|unweighted",
"table4|no_oil_transition|2007-2022|unweighted",
"table5|percapita_post|unweighted",
"table6|avg3_2005-07_to_2020-22|unweighted",
"table6|end2019_2007-2019|unweighted"
],
"distinct_pass_wild_10pct_full": [
"table4|core_no_norway|2007-2017|unweighted",
"table4|net_importers_1990|1997-2007|unweighted",
"table4|net_importers_1997|1997-2007|unweighted",
"table4|no_oil_transition|2007-2017|unweighted"
],
"distinct_pass_wild_10pct_after_screen": [
"table4|all|2007-2017|unweighted",
"table4|all|2007-2017|weighted",
"table4|net_importers_1990|1997-2007|unweighted",
"table4|net_importers_1997|1990-2007|unweighted",
"table4|net_importers_1997|1990-2007|weighted",
"table4|net_importers_1997|2007-2017|unweighted",
"table5|start1995_pre|unweighted",
"table6|avg3_2005-07_to_2020-22|weighted"
],
"malta_largest_share_hc_variance": [
"table4|all|2007-2017|unweighted",
"table4|all|2007-2022|unweighted",
"table4|core_no_norway|2007-2017|unweighted",
"table4|core_no_norway|2007-2022|unweighted",
"table4|net_importers_1997|2007-2017|unweighted",
"table4|net_importers_1997|2007-2022|unweighted",
"table4|no_oil_transition|2007-2017|unweighted",
"table4|no_oil_transition|2007-2022|unweighted",
"table5|percapita_post|unweighted",
"table6|avg3_2005-07_to_2020-22|unweighted",
"table6|benchmark_2007-2017|unweighted",
"table6|end2019_2007-2019|unweighted"
],
"malta_share_hc_variance_over_half": [
"table4|all|2007-2017|unweighted",
"table4|all|2007-2022|unweighted",
"table4|core_no_norway|2007-2017|unweighted",
"table4|core_no_norway|2007-2022|unweighted",
"table4|net_importers_1997|2007-2017|unweighted",
"table4|net_importers_1997|2007-2022|unweighted",
"table4|no_oil_transition|2007-2017|unweighted",
"table4|no_oil_transition|2007-2022|unweighted",
"table5|percapita_post|unweighted",
"table6|avg3_2005-07_to_2020-22|unweighted",
"table6|benchmark_2007-2017|unweighted",
"table6|end2019_2007-2019|unweighted"
],
"malta_cooks_d_over_4_over_n": [
"table4|all|2007-2017|unweighted",
"table4|all|2007-2022|unweighted",
"table4|core_no_norway|2007-2017|unweighted",
"table4|core_no_norway|2007-2022|unweighted",
"table4|net_importers_1997|2007-2017|unweighted",
"table4|net_importers_1997|2007-2022|unweighted",
"table4|no_oil_transition|2007-2017|unweighted",
"table4|no_oil_transition|2007-2022|unweighted",
"table5|percapita_post|unweighted",
"table6|avg3_2005-07_to_2020-22|unweighted",
"table6|benchmark_2007-2017|unweighted",
"table6|end2019_2007-2019|unweighted"
],
"unweighted_malta_share_range": [
0,
0.867
],
"weighted_malta_share_max": 0.0238,
"pass_wild_10pct_full": [
"table4|core_no_norway|2007-2017|unweighted",
"table4|net_importers_1990|1997-2007|unweighted",
"table4|net_importers_1997|1997-2007|unweighted",
"table4|no_oil_transition|2007-2017|unweighted"
],
"pass_wild_10pct_without_malta": [
"table4|core_no_norway|2007-2017|unweighted",
"table4|net_importers_1997|1997-2007|unweighted",
"table4|no_oil_transition|2007-2017|unweighted"
],
"pass_hc3_10pct_without_malta": [
"table4|all|2017-2022|weighted",
"table4|core_no_norway|2007-2017|unweighted",
"table4|no_oil_transition|2007-2017|unweighted"
],
"pass_wild_10pct_after_screen": [
"table4|all|2007-2017|unweighted",
"table4|all|2007-2017|weighted",
"table4|net_importers_1990|1997-2007|unweighted",
"table4|net_importers_1997|1990-2007|unweighted",
"table4|net_importers_1997|1990-2007|weighted",
"table4|net_importers_1997|2007-2017|unweighted",
"table5|start1995_pre|unweighted",
"table6|avg3_2005-07_to_2020-22|weighted",
"table6|benchmark_2007-2017|unweighted",
"table6|benchmark_2007-2017|weighted"
]
},
"multiple_testing_table4_unweighted_without_malta": {
"n_tests": 20,
"n_pass_10pct": 3,
"expected_pass_under_null_10pct": 2,
"prob_ge_observed_if_independent": 0.3231,
"min_p": 0.0436,
"bonferroni_threshold_10pct": 0.005,
"bonferroni_threshold_5pct": 0.0025,
"bonferroni_min_adjusted_p": 0.872,
"holm_min_adjusted_p": 0.872,
"n_reject_bonferroni_10pct": 0,
"n_reject_holm_10pct": 0,
"n_reject_holm_5pct": 0,
"labels": [
"all|1990-2007",
"all|2007-2022",
"all|1997-2007",
"all|2007-2017",
"all|2017-2022",
"no_oil_transition|1990-2007",
"no_oil_transition|2007-2022",
"no_oil_transition|1997-2007",
"no_oil_transition|2007-2017",
"no_oil_transition|2017-2022",
"net_importers_1997|1990-2007",
"net_importers_1997|2007-2022",
"net_importers_1997|1997-2007",
"net_importers_1997|2007-2017",
"net_importers_1997|2017-2022",
"core_no_norway|2007-2022",
"core_no_norway|2007-2017",
"core_no_norway|2017-2022",
"net_importers_1990|1990-2007",
"net_importers_1990|1997-2007"
]
},
"note": "Malta deletions use the seed of the full-sample row; diagnostics are Malta's in the full sample (weighted rows: WLS with base-year co2 weights); the Cook's-distance screen drops every country with D > 4/n in each sample, once, and is a sensitivity check only. All round-7 quantities were added after the main results were known."
}
},
"test": {
"high_wedge_pre_positive_5pct": false,
"high_wedge_post_smaller": false,
"high_wedge_change_p_stacked": 0.5201
},
"hypothesis_supported": false
}Experiment script · experiment_v1.py
"""Decoupling or displacement? GDP elasticities of consumption-based vs territorial CO2, 1990-2022.
Real data: the user-supplied OWID CO2 file, loaded via dataset.load(). Nothing is simulated.
"""
import json
import warnings
import numpy as np
import pandas as pd
import statsmodels.formula.api as smf
from dataset import load
warnings.simplefilter("ignore")
np.random.seed(0)
df = load()
NEED = ["country", "iso_code", "year", "gdp", "population", "co2", "consumption_co2", "trade_co2", "trade_co2_share"]
missing_cols = [c for c in NEED if c not in df.columns]
print("rows, cols:", df.shape, "| missing required columns:", missing_cols)
for c in NEED[3:]:
df[c] = pd.to_numeric(df[c], errors="coerce")
# ---------------------------------------------------------------- sample
s = df[df["iso_code"].notna()].copy()
n_ent = s["country"].nunique()
s = s[(s["year"] >= 1990) & (s["year"] <= 2022)]
s = s.dropna(subset=["gdp", "population", "co2", "consumption_co2"])
s = s[(s["co2"] > 0) & (s["consumption_co2"] > 0) & (s["gdp"] > 0)]
kuw = (s["country"] == "Kuwait") & (s["year"] <= 1992)
s = s[~kuw].sort_values(["country", "year"]).reset_index(drop=True)
print(f"ISO entities {n_ent}; analysis sample {s['country'].nunique()} countries, {len(s)} country-years; Kuwait rows dropped {int(kuw.sum())}")
for v in ["gdp", "co2", "consumption_co2", "population"]:
s["l_" + v] = np.log(s[v])
s["l_wedge"] = s["l_consumption_co2"] - s["l_co2"]
s["l_gdppc"] = s["l_gdp"] - s["l_population"]
acc = (s["consumption_co2"] - s["co2"] - s["trade_co2"]).abs()
print("accounting check max|cons - co2 - trade_co2| (Mt):", float(acc.max()))
# baseline (1990) income terciles from gdp/population, taken from the unfiltered country rows
b = df[(df["iso_code"].notna()) & (df["year"] == 1990)].set_index("country")
gpc90 = (b["gdp"] / b["population"]).reindex(s["country"].unique()).dropna()
q1, q2 = gpc90.quantile([1 / 3, 2 / 3])
terc = pd.Series(np.where(gpc90 >= q2, "high", np.where(gpc90 >= q1, "mid", "low")), index=gpc90.index)
pop90 = b["population"]
s["terc"] = s["country"].map(terc)
print("tercile cut-offs (1990 int-$ 2011 per person):", round(float(q1)), round(float(q2)), terc.value_counts().to_dict())
def star(p):
return "***" if p < 0.01 else "**" if p < 0.05 else "*" if p < 0.1 else ""
def coef(m, name):
b_, se, p = float(m.params[name]), float(m.bse[name]), float(m.pvalues[name])
return {"b": round(b_, 4), "se": round(se, 4), "p": round(p, 4), "stars": star(p)}
# ---------------------------------------------------------------- (1) long differences
def long_diff(a, z, frame):
A = frame[frame["year"] == a].set_index("country")
Z = frame[frame["year"] == z].set_index("country")
j = A.index.intersection(Z.index)
out = pd.DataFrame({k: Z.loc[j, "l_" + k] - A.loc[j, "l_" + k] for k in ["gdp", "co2", "consumption_co2", "wedge", "gdppc"]})
out["terc"] = terc.reindex(j)
out["pop90"] = pop90.reindex(j)
out["w0"] = A.loc[j, "co2"]
out["period"] = f"{a}-{z}"
return out.dropna(subset=["terc"]).reset_index().rename(columns={"index": "country"})
def ld_block(d):
res = {"n": int(len(d))}
for y in ["co2", "consumption_co2", "wedge"]:
m = smf.ols(f"{y} ~ 0 + C(terc) + C(terc):gdp", d).fit(cov_type="HC1")
res[y] = {t: coef(m, f"C(terc)[{t}]:gdp") for t in ["low", "mid", "high"]}
res[y]["r2"] = round(float(m.rsquared), 4)
m = smf.ols("wedge ~ 0 + C(terc) + C(terc):gdp", d).fit(cov_type="HC1")
t = m.t_test("C(terc)[high]:gdp - C(terc)[low]:gdp = 0")
res["wedge_high_minus_low"] = {"b": round(float(t.effect[0]), 4), "se": round(float(t.sd[0][0]), 4), "p": round(float(t.pvalue), 4)}
means = d.groupby("terc")[["gdp", "co2", "consumption_co2", "wedge"]].mean().round(4)
res["mean_dlog"] = {k: {kk: float(vv) for kk, vv in v.items()} for k, v in means.to_dict("index").items()}
res["mean_wedge_test"] = {}
for t_ in ["low", "mid", "high"]:
w = d.loc[d["terc"] == t_, "wedge"]
m0 = smf.ols("wedge ~ 1", d[d["terc"] == t_]).fit(cov_type="HC1")
res["mean_wedge_test"][t_] = {"mean": round(float(w.mean()), 4), "median": round(float(w.median()), 4),
"se": round(float(m0.bse["Intercept"]), 4), "p": round(float(m0.pvalues["Intercept"]), 4)}
hi = d[d["terc"] == "high"]
res["high_dlog_co2_minus_cons"] = round(float(hi["co2"].mean() - hi["consumption_co2"].mean()), 4)
mw = smf.wls("wedge ~ gdp", hi, weights=hi["w0"]).fit(cov_type="HC1")
res["high_wedge_wls_baseline_co2"] = coef(mw, "gdp")
res["high_abs_decoupled_territorial"] = int(((hi["co2"] < 0) & (hi["gdp"] > 0)).sum())
res["high_abs_decoupled_consumption"] = int(((hi["consumption_co2"] < 0) & (hi["gdp"] > 0)).sum())
res["high_n"] = int(len(hi))
return res
def stacked(d1, d2):
d = pd.concat([d1, d2], ignore_index=True)
d["post"] = (d["period"] == d2["period"].iloc[0]).astype(int)
m = smf.ols("wedge ~ 0 + C(terc):C(post) + C(terc):C(post):gdp", d).fit(
cov_type="cluster", cov_kwds={"groups": pd.factorize(d["country"])[0]})
t = m.t_test("C(terc)[high]:C(post)[1]:gdp - C(terc)[high]:C(post)[0]:gdp = 0")
return {"change_high_wedge": round(float(t.effect[0]), 4), "se": round(float(t.sd[0][0]), 4),
"p": round(float(t.pvalue), 4), "n_obs": int(len(d)), "n_countries": int(d["country"].nunique())}
pre, post = long_diff(1990, 2007, s), long_diff(2007, 2022, s)
LD = {"1990-2007": ld_block(pre), "2007-2022": ld_block(post), "stacked": stacked(pre, post)}
print("LD wedge high pre/post:", LD["1990-2007"]["wedge"]["high"], LD["2007-2022"]["wedge"]["high"], LD["stacked"])
# ---------------------------------------------------------------- (2) annual first differences
g = s.groupby("country")
consec = g["year"].diff() == 1
for v in ["gdp", "co2", "consumption_co2", "wedge"]:
s["d_" + v] = g["l_" + v].diff().where(consec)
fd = s.dropna(subset=["d_gdp", "d_co2", "d_consumption_co2", "terc"]).copy()
fd["post"] = (fd["year"] >= 2008).astype(int)
fd["tcode"] = fd["year"] - fd["year"].min()
FD = {"n_obs": int(len(fd)), "n_countries": int(fd["country"].nunique())}
for y in ["co2", "consumption_co2", "wedge"]:
f = f"d_{y} ~ C(year) + C(terc):C(post):d_gdp"
mdk = smf.ols(f, fd).fit(cov_type="hac-groupsum", cov_kwds={"time": fd["tcode"].values, "maxlags": 2})
mcl = smf.ols(f, fd).fit(cov_type="cluster", cov_kwds={"groups": pd.factorize(fd["country"])[0]})
FD[y] = {"r2": round(float(mdk.rsquared), 4)}
for t_ in ["low", "mid", "high"]:
for p_ in [0, 1]:
nm = f"C(terc)[{t_}]:C(post)[{p_}]:d_gdp"
FD[y][f"{t_}_{'post' if p_ else 'pre'}"] = {"dk": coef(mdk, nm), "cl_se": round(float(mcl.bse[nm]), 4)}
if y == "wedge":
t = mdk.t_test("C(terc)[high]:C(post)[1]:d_gdp - C(terc)[high]:C(post)[0]:d_gdp = 0")
FD["wedge_high_change_dk"] = {"b": round(float(t.effect[0]), 4), "se": round(float(t.sd[0][0]), 4), "p": round(float(t.pvalue), 4)}
print("FD wedge high pre/post:", FD["wedge"]["high_pre"], FD["wedge"]["high_post"])
# ---------------------------------------------------------------- (3) trade shares, rich tercile
shares = {}
for yr in [1990, 2007, 2022]:
r = s[(s["year"] == yr) & (s["terc"] == "high")]
shares[str(yr)] = {"mean_trade_co2_share": round(float(r["trade_co2_share"].mean()), 3),
"median_trade_co2_share": round(float(r["trade_co2_share"].median()), 3),
"aggregate_trade_over_co2_pct": round(float(100 * r["trade_co2"].sum() / r["co2"].sum()), 3),
"aggregate_trade_co2_mt": round(float(r["trade_co2"].sum()), 1),
"n": int(len(r)), "share_net_importers": round(float((r["trade_co2"] > 0).mean()), 3)}
low22 = s[(s["year"] == 2022) & (s["terc"] == "low")]
shares["2022_low_mean_trade_co2_share"] = round(float(low22["trade_co2_share"].mean()), 3)
# ---------------------------------------------------------------- (4) robustness
ROB = {}
p95 = long_diff(1995, 2007, s)
m = smf.ols("wedge ~ 0 + C(terc) + C(terc):gdp", p95).fit(cov_type="HC1")
ROB["start1995_high_wedge_pre"] = coef(m, "C(terc)[high]:gdp") | {"n": int(len(p95))}
for lab, d in [("pre", pre), ("post", post)]:
big = d[d["pop90"] >= 1e6]
m = smf.ols("wedge ~ 0 + C(terc) + C(terc):gdp", big).fit(cov_type="HC1")
ROB[f"pop1m_high_wedge_{lab}"] = coef(m, "C(terc)[high]:gdp") | {"n": int(len(big))}
m = smf.ols("wedge ~ 0 + C(terc) + C(terc):gdppc", d).fit(cov_type="HC1")
ROB[f"percapita_high_wedge_{lab}"] = coef(m, "C(terc)[high]:gdppc")
lv = s.dropna(subset=["terc"]).copy()
lv["post"] = (lv["year"] >= 2008).astype(int)
m = smf.ols("l_wedge ~ C(country) + C(year) + C(terc):C(post):l_gdp", lv).fit(
cov_type="cluster", cov_kwds={"groups": pd.factorize(lv["country"])[0]})
ROB["fe_levels_high_pre"] = coef(m, "C(terc)[high]:C(post)[0]:l_gdp")
ROB["fe_levels_high_post"] = coef(m, "C(terc)[high]:C(post)[1]:l_gdp")
ROB["fe_levels_n"] = int(len(lv))
print("robustness:", json.dumps(ROB))
# ================================================================ REVISION (referee round 1)
# Everything above is unchanged, so every earlier number reproduces. No randomness is used below.
def ci(c):
return [round(c["b"] - 1.959964 * c["se"], 4), round(c["b"] + 1.959964 * c["se"], 4)]
def make_sample(kmode):
t = df[df["iso_code"].notna()]
t = t[(t["year"] >= 1990) & (t["year"] <= 2022)].dropna(subset=["gdp", "population", "co2", "consumption_co2"])
t = t[(t["co2"] > 0) & (t["consumption_co2"] > 0) & (t["gdp"] > 0)]
k = t["country"] == "Kuwait"
if kmode == "drop1990_92":
t = t[~(k & (t["year"] <= 1992))]
elif kmode == "drop1991_92":
t = t[~(k & t["year"].isin([1991, 1992]))]
t = t.sort_values(["country", "year"]).reset_index(drop=True)
for v in ["gdp", "co2", "consumption_co2", "population"]:
t["l_" + v] = np.log(t[v])
t["l_wedge"] = t["l_consumption_co2"] - t["l_co2"]
t["l_gdppc"] = t["l_gdp"] - t["l_population"]
t["terc"] = t["country"].map(terc)
return t
def fd_frame(t):
t = t.copy()
gg = t.groupby("country")
cons_ = gg["year"].diff() == 1
for v in ["gdp", "co2", "consumption_co2", "wedge"]:
t["d_" + v] = gg["l_" + v].diff().where(cons_)
f_ = t.dropna(subset=["d_gdp", "d_co2", "d_consumption_co2", "terc"]).copy()
f_["post"] = (f_["year"] >= 2008).astype(int)
f_["tcode"] = f_["year"] - f_["year"].min()
return f_
def fd_high_wedge(f_, cl=True):
fml = "d_wedge ~ C(year) + C(terc):C(post):d_gdp"
mdk = smf.ols(fml, f_).fit(cov_type="hac-groupsum", cov_kwds={"time": f_["tcode"].values, "maxlags": 2})
out = {}
for p_, lab in [(0, "pre"), (1, "post")]:
nm = f"C(terc)[high]:C(post)[{p_}]:d_gdp"
out[lab] = coef(mdk, nm)
if cl:
mcl = smf.ols(fml, f_).fit(cov_type="cluster", cov_kwds={"groups": pd.factorize(f_["country"])[0]})
for p_, lab in [(0, "pre"), (1, "post")]:
out[lab]["cl_se"] = round(float(mcl.bse[f"C(terc)[high]:C(post)[{p_}]:d_gdp"]), 4)
return out
def high_slope(d, y="wedge", x="gdp"):
return coef(smf.ols(f"{y} ~ 0 + C(terc) + C(terc):{x}", d).fit(cov_type="HC1"), f"C(terc)[high]:{x}")
def high_wls(d):
hi = d[d["terc"] == "high"]
return coef(smf.wls("wedge ~ gdp", hi, weights=hi["w0"]).fit(cov_type="HC1"), "gdp")
def key_estimates(t):
a, z = long_diff(1990, 2007, t), long_diff(2007, 2022, t)
out = {"n_countries": int(t["country"].nunique()), "n_obs": int(len(t)), "ld_n": [int(len(a)), int(len(z))]}
for y in ["co2", "consumption_co2", "wedge"]:
out[f"ld_high_{y}"] = {"pre": high_slope(a, y), "post": high_slope(z, y)}
out["ld_stacked_change"] = stacked(a, z)
out["ld_weighted"] = {"pre": high_wls(a), "post": high_wls(z)}
out["fd_high_wedge"] = fd_high_wedge(fd_frame(t))
out["start1995_pre"] = high_slope(long_diff(1995, 2007, t))
out["pop1m"] = {"pre": high_slope(a[a["pop90"] >= 1e6]), "post": high_slope(z[z["pop90"] >= 1e6])}
out["percapita"] = {"pre": high_slope(a, x="gdppc"), "post": high_slope(z, x="gdppc")}
lv_ = t.dropna(subset=["terc"]).copy()
lv_["post"] = (lv_["year"] >= 2008).astype(int)
mm = smf.ols("l_wedge ~ C(country) + C(year) + C(terc):C(post):l_gdp", lv_).fit(
cov_type="cluster", cov_kwds={"groups": pd.factorize(lv_["country"])[0]})
out["fe_levels"] = {"pre": coef(mm, "C(terc)[high]:C(post)[0]:l_gdp"), "post": coef(mm, "C(terc)[high]:C(post)[1]:l_gdp")}
return out
# --- change 1/2: Kuwait facts, exploratory replication, Kuwait variants
kw = df[df["country"] == "Kuwait"].set_index("year")
kof = df[df["country"] == "Kuwaiti Oil Fires"]
KUWAIT = {"co2_1990": round(float(kw.loc[1990, "co2"]), 2), "co2_1991": round(float(kw.loc[1991, "co2"]), 2),
"co2_1992": round(float(kw.loc[1992, "co2"]), 2),
"gdp_change_1990_91_pct": round(float(100 * (kw.loc[1991, "gdp"] / kw.loc[1990, "gdp"] - 1)), 1),
"oil_fires_entity_nonmissing_co2": int(kof["co2"].notna().sum()), "tercile": str(terc.get("Kuwait"))}
# exact replication of the exploratory specification run before the plan was fixed (Kuwait 1990-1992 kept,
# rich = top tercile vs rest, pooled 1991-2022, year FE, differences without a consecutive-year check, clustered by country)
ex = df[df["iso_code"].notna() & df["year"].between(1990, 2022)].dropna(subset=["gdp", "co2", "consumption_co2", "population"])
ex = ex[(ex["co2"] > 0) & (ex["consumption_co2"] > 0)].sort_values(["country", "year"]).copy()
for v in ["gdp", "co2", "consumption_co2"]:
ex["d" + v] = ex.groupby("country")[v].transform(lambda z_: np.log(z_).diff())
ex["dwedge"] = ex["dconsumption_co2"] - ex["dco2"]
g90 = ex[ex["year"] == 1990].set_index("country")
g90 = g90["gdp"] / g90["population"]
ex["rich"] = (ex["country"].map(g90) >= g90.quantile(2 / 3)).astype(int)
ex = ex.dropna(subset=["dgdp", "dwedge"]).loc[lambda z_: z_["country"].isin(g90.index)]
mx = smf.ols("dwedge ~ dgdp*rich + C(year)", ex).fit(cov_type="cluster", cov_kwds={"groups": ex["country"].astype("category").cat.codes})
EXPLORATORY = {"rich_x_dgdp": coef(mx, "dgdp:rich"), "dgdp_rest": coef(mx, "dgdp"),
"rich_total": round(float(mx.params["dgdp"] + mx.params["dgdp:rich"]), 4), "n_obs": int(len(ex))}
mx2 = smf.ols("dwedge ~ dgdp*rich + C(year)", ex[ex["country"] != "Kuwait"]).fit(
cov_type="cluster", cov_kwds={"groups": ex.loc[ex["country"] != "Kuwait", "country"].astype("category").cat.codes})
EXPLORATORY["rich_x_dgdp_without_kuwait"] = coef(mx2, "dgdp:rich")
print("exploratory replication:", EXPLORATORY)
KEY = {m_: key_estimates(make_sample(m_)) for m_ in ["drop1990_92", "keep_all", "drop1991_92"]}
assert abs(KEY["drop1990_92"]["ld_high_wedge"]["pre"]["b"] - LD["1990-2007"]["wedge"]["high"]["b"]) < 1e-9
print("Kuwait variants, high wedge LD pre/post, FD pre/post:",
{k: (v["ld_high_wedge"]["pre"]["b"], v["ld_high_wedge"]["post"]["b"], v["fd_high_wedge"]["pre"]["b"], v["fd_high_wedge"]["post"]["b"]) for k, v in KEY.items()})
# leave-one-country-out ranges for the headline high-tercile wedge elasticities (baseline sample)
fd0 = fd_frame(s)
loo_rows = []
for c_ in sorted(s["country"].unique()):
a_, z_ = pre[pre["country"] != c_], post[post["country"] != c_]
f_ = fd_high_wedge(fd0[fd0["country"] != c_], cl=False)
loo_rows.append({"country": c_, "ld_pre": high_slope(a_)["b"], "ld_post": high_slope(z_)["b"],
"wls_pre": high_wls(a_)["b"], "wls_post": high_wls(z_)["b"],
"fd_pre": f_["pre"]["b"], "fd_post": f_["post"]["b"]})
loo = pd.DataFrame(loo_rows).set_index("country")
LOO = {k: {"min": round(float(loo[k].min()), 4), "argmin": str(loo[k].idxmin()), "max": round(float(loo[k].max()), 4),
"argmax": str(loo[k].idxmax())} for k in loo.columns}
LOO["n_fits"] = int(len(loo))
print("LOO:", LOO)
# --- change 4: tercile intercepts (drift) with growth demeaned within tercile; raw intercepts; growth share
def intercepts(d):
d = d.copy()
d["gdp_dm"] = d["gdp"] - d.groupby("terc")["gdp"].transform("mean")
mdm = smf.ols("wedge ~ 0 + C(terc) + C(terc):gdp_dm", d).fit(cov_type="HC1")
mraw = smf.ols("wedge ~ 0 + C(terc) + C(terc):gdp", d).fit(cov_type="HC1")
out = {}
for t_ in ["low", "mid", "high"]:
mg = float(d.loc[d["terc"] == t_, "gdp"].mean())
slope = float(mraw.params[f"C(terc)[{t_}]:gdp"])
out[t_] = {"intercept_demeaned": coef(mdm, f"C(terc)[{t_}]"), "intercept_raw": coef(mraw, f"C(terc)[{t_}]"),
"mean_dlog_gdp": round(mg, 4), "growth_component": round(slope * mg, 4),
"growth_share_of_mean": round(slope * mg / float(mdm.params[f"C(terc)[{t_}]"]), 4)}
hi = d[d["terc"] == "high"].copy()
wm = float(np.average(hi["gdp"], weights=hi["w0"]))
hi["gdp_wdm"] = hi["gdp"] - wm
mw = smf.wls("wedge ~ gdp_wdm", hi, weights=hi["w0"]).fit(cov_type="HC1")
mwr = smf.wls("wedge ~ gdp", hi, weights=hi["w0"]).fit(cov_type="HC1")
out["high_weighted"] = {"intercept_demeaned": coef(mw, "Intercept"), "intercept_raw": coef(mwr, "Intercept"),
"slope": coef(mw, "gdp_wdm"), "weighted_mean_dlog_gdp": round(wm, 4)}
return out
INTER = {"1990-2007": intercepts(pre), "2007-2022": intercepts(post)}
def agg_wedge(a, z, frame, members):
A = frame[(frame["year"] == a) & frame["country"].isin(members)].set_index("country")
Z = frame[(frame["year"] == z) & frame["country"].isin(members)].set_index("country")
j = A.index.intersection(Z.index)
wa = np.log(A.loc[j, "consumption_co2"].sum() / A.loc[j, "co2"].sum())
wz = np.log(Z.loc[j, "consumption_co2"].sum() / Z.loc[j, "co2"].sum())
return {"n": int(len(j)), "w_start": round(float(wa), 4), "w_end": round(float(wz), 4), "change": round(float(wz - wa), 4)}
highs = terc[terc == "high"].index
AGG = {"balanced_1990_2007": agg_wedge(1990, 2007, s, highs), "balanced_2007_2022": agg_wedge(2007, 2022, s, highs),
"table1_implied_1990_2007": round(float(np.log((100 + shares["2007"]["aggregate_trade_over_co2_pct"]) / (100 + shares["1990"]["aggregate_trade_over_co2_pct"]))), 4),
"table1_implied_2007_2022": round(float(np.log((100 + shares["2022"]["aggregate_trade_over_co2_pct"]) / (100 + shares["2007"]["aggregate_trade_over_co2_pct"]))), 4)}
# --- change 7: endpoint averaging, alternative end year, benchmark-to-benchmark differences
def long_diff_avg(wa, wz, frame):
def win(w):
x = frame[frame["year"].between(w[0], w[1])]
cnt = x.groupby("country")["year"].count()
m_ = x.groupby("country")[["l_gdp", "l_co2", "l_consumption_co2", "l_wedge", "l_gdppc", "co2"]].mean()
return m_[cnt.reindex(m_.index) == (w[1] - w[0] + 1)]
A, Z = win(wa), win(wz)
j = A.index.intersection(Z.index)
out = pd.DataFrame({k: Z.loc[j, "l_" + k] - A.loc[j, "l_" + k] for k in ["gdp", "co2", "consumption_co2", "wedge", "gdppc"]})
out["terc"], out["w0"] = terc.reindex(j), A.loc[j, "co2"]
return out.dropna(subset=["terc"])
def ep_block(d):
hi = d[d["terc"] == "high"]
return {"n": int(len(d)), "high_n": int(len(hi)), "co2": high_slope(d, "co2"), "consumption_co2": high_slope(d, "consumption_co2"),
"wedge": high_slope(d), "wedge_weighted": high_wls(d), "high_mean_wedge": round(float(hi["wedge"].mean()), 4),
"high_median_wedge": round(float(hi["wedge"].median()), 4)}
ENDPOINTS = {"avg3_1990-92_to_2005-07": ep_block(long_diff_avg((1990, 1992), (2005, 2007), s)),
"avg3_2005-07_to_2020-22": ep_block(long_diff_avg((2005, 2007), (2020, 2022), s)),
"end2019_2007-2019": ep_block(long_diff(2007, 2019, s)),
"benchmark_1997-2007": ep_block(long_diff(1997, 2007, s)),
"benchmark_2007-2017": ep_block(long_diff(2007, 2017, s))}
# Status of consumption_co2 by year, per the Global Carbon Budget 2025 supplement (Sect. S1): detailed MRIO estimates
# for the GTAP years below, extended to an annual series 1990-latest GDP year with GDP and trade time series.
BENCH = [1997, 2001, 2004, 2007, 2011, 2014, 2017]
def status(y):
return "benchmark" if y in BENCH else "extended_back" if y < 1997 else "extended_forward" if y > 2017 else "between_benchmarks"
ENDPOINT_STATUS = {str(y): status(y) for y in [1990, 1992, 1995, 1997, 2005, 2006, 2007, 2017, 2019, 2020, 2022]}
print("endpoints:", {k: (v["wedge"]["b"], v["wedge"]["se"], v["wedge_weighted"]["b"]) for k, v in ENDPOINTS.items()})
REVISION = {"kuwait_facts": KUWAIT, "exploratory_replication": EXPLORATORY, "kuwait_variants": KEY, "leave_one_out": LOO,
"intercepts": INTER, "aggregate_wedge_high": AGG, "endpoints": ENDPOINTS, "consumption_status_gcb2025": ENDPOINT_STATUS,
"ci95": {"ld_high_wedge_pre": ci(LD["1990-2007"]["wedge"]["high"]), "ld_high_wedge_post": ci(LD["2007-2022"]["wedge"]["high"]),
"stacked_change": ci({"b": LD["stacked"]["change_high_wedge"], "se": LD["stacked"]["se"]}),
"weighted_pre": ci(LD["1990-2007"]["high_wedge_wls_baseline_co2"]), "weighted_post": ci(LD["2007-2022"]["high_wedge_wls_baseline_co2"]),
"fe_levels_pre": ci(ROB["fe_levels_high_pre"]), "fe_levels_post": ci(ROB["fe_levels_high_post"])}}
# ================================================================ REVISION (referee round 2)
# Drift on benchmark-year and alternative endpoints for the top tercile (no randomness).
def drift_block(a, z):
d = long_diff(a, z, s)
it = intercepts(d)
hi = d[d["terc"] == "high"]
return {"n_high": int(len(hi)), "mean_wedge": round(float(hi["wedge"].mean()), 4), "median_wedge": round(float(hi["wedge"].median()), 4),
"mean_dlog_gdp": round(float(hi["gdp"].mean()), 4), "mean_dlog_co2": round(float(hi["co2"].mean()), 4),
"mean_dlog_cons": round(float(hi["consumption_co2"].mean()), 4),
"intercept_demeaned": it["high"]["intercept_demeaned"], "intercept_raw": it["high"]["intercept_raw"],
"weighted_intercept_demeaned": it["high_weighted"]["intercept_demeaned"], "weighted_intercept_raw": it["high_weighted"]["intercept_raw"],
"aggregate_balanced": agg_wedge(a, z, s, highs)}
ROUND2 = {"1997-2007": drift_block(1997, 2007), "2007-2017": drift_block(2007, 2017), "2007-2019": drift_block(2007, 2019),
"2017-2022": drift_block(2017, 2022), "2007-2022_check": drift_block(2007, 2022)}
assert abs(ROUND2["2007-2022_check"]["aggregate_balanced"]["change"] - AGG["balanced_2007_2022"]["change"]) < 1e-12
REVISION["round2_drift"] = ROUND2
print("round2 drift:", {k: (v["mean_wedge"], v["median_wedge"], v["weighted_intercept_demeaned"]["b"], v["aggregate_balanced"]["change"]) for k, v in ROUND2.items()})
# ================================================================ REVISION (referee round 3)
# Everything above is unchanged. New randomness uses its own seeded generators (np.random.default_rng), so the
# global numpy stream seeded above is never touched and every earlier number reproduces.
import statsmodels.api as sm
R3 = {}
# --- change 1: group membership and composition variants of the top tercile
R3["members"] = {t_: sorted(terc[terc == t_].index.tolist()) for t_ in ["low", "mid", "high"]}
OIL = ["Kuwait", "Venezuela", "Trinidad and Tobago", "United Arab Emirates", "Saudi Arabia"]
TRANS = ["Slovakia", "Lithuania", "Czechia", "Latvia", "Estonia", "Slovenia"]
assert all(terc.get(c_) == "high" for c_ in OIL + TRANS)
t97 = s[(s["year"] == 1997) & (s["terc"] == "high")].set_index("country")["trade_co2"]
VARIANTS = {"all": list(highs), "no_oil_transition": [c_ for c_ in highs if c_ not in OIL + TRANS],
"net_importers_1997": sorted(t97[t97 > 0].index.tolist())}
R3["variant_members"] = {k: sorted(v) for k, v in VARIANTS.items()}
R3["high_net_exporters_1997"] = sorted(t97[t97 <= 0].index.tolist())
R3["high_net_exporters_2007"] = sorted(s[(s["year"] == 2007) & (s["terc"] == "high") & (s["trade_co2"] <= 0)]["country"].tolist())
def keep_high(d, members):
return d[(d["terc"] != "high") | d["country"].isin(members)].copy()
def hc3_slope(y, x, w=None):
X = sm.add_constant(np.asarray(x, float))
m = (sm.WLS(np.asarray(y, float), X, weights=np.asarray(w, float)) if w is not None else sm.OLS(np.asarray(y, float), X)).fit(cov_type="HC3")
b_, se = float(m.params[1]), float(m.bse[1])
return {"b": round(b_, 4), "se_hc3": round(se, 4), "p_hc3": round(float(m.pvalues[1]), 4),
"ci95_hc3": [round(b_ - 1.959964 * se, 4), round(b_ + 1.959964 * se, 4)]}
WEBB = np.array([-np.sqrt(1.5), -1.0, -np.sqrt(0.5), np.sqrt(0.5), 1.0, np.sqrt(1.5)])
def wild_p(y, x, w, seed, B=9999):
"""Wild bootstrap (Webb weights) p-value for slope = 0, null imposed, HC3 t-statistic; WLS if w given."""
y, x = np.asarray(y, float), np.asarray(x, float)
sw = np.sqrt(np.ones_like(y) if w is None else np.asarray(w, float) / np.mean(w))
X = np.column_stack([np.ones_like(x), x]) * sw[:, None]
Y = y * sw
XtXi = np.linalg.inv(X.T @ X)
h = np.einsum("ij,jk,ik->i", X, XtXi, X)
def tstat(Ymat): # Ymat: n x B
beta = XtXi @ X.T @ Ymat
e = Ymat - X @ beta
u = e / (1 - h)[:, None]
v = np.einsum("j,jb->b", (XtXi @ X.T)[1] ** 2, u ** 2)
return beta[1] / np.sqrt(v)
t0 = float(tstat(Y[:, None])[0])
fit0 = sw * (np.sum(sw * Y) / np.sum(sw * sw)) # weighted mean under the null, in transformed units
e0 = Y - fit0
rng = np.random.default_rng(seed)
V = WEBB[rng.integers(0, 6, size=(len(y), B))]
tb = tstat(fit0[:, None] + e0[:, None] * V)
return round(float(np.mean(np.abs(tb) >= abs(t0))), 4)
def kish(w):
w = np.asarray(w, float)
return round(float(w.sum() ** 2 / (w ** 2).sum()), 2)
def group_block(d, members, a, z, seed):
dv = keep_high(d, members)
hi = dv[dv["terc"] == "high"]
out = {"n_high": int(len(hi)), "n_total": int(len(dv))}
for y in ["co2", "consumption_co2", "wedge"]:
c = high_slope(dv, y)
out[f"slope_{y}_hc1"] = c | {"ci95": ci(c)}
out["slope_wedge_hc3"] = hc3_slope(hi["wedge"], hi["gdp"])
out["slope_wedge_wild_p"] = wild_p(hi["wedge"], hi["gdp"], None, seed)
out["slope_wedge_weighted_hc1"] = high_wls(dv)
out["slope_wedge_weighted_hc3"] = hc3_slope(hi["wedge"], hi["gdp"], hi["w0"])
out["slope_wedge_weighted_wild_p"] = wild_p(hi["wedge"], hi["gdp"], hi["w0"], seed + 1)
out["kish_n_weighted"] = kish(hi["w0"])
it = intercepts(dv)
out["mean_dlog_gdp"] = round(float(hi["gdp"].mean()), 4)
out["mean_wedge"], out["median_wedge"] = round(float(hi["wedge"].mean()), 4), round(float(hi["wedge"].median()), 4)
out["intercept_demeaned"], out["intercept_raw"] = it["high"]["intercept_demeaned"], it["high"]["intercept_raw"]
out["weighted_intercept_demeaned"] = it["high_weighted"]["intercept_demeaned"]
out["aggregate_balanced"] = agg_wedge(a, z, s, members)
sl = out["slope_wedge_hc1"]
mg, mw = out["mean_dlog_gdp"], out["intercept_demeaned"]["b"]
out["upper_ci_times_mean_growth"] = round(sl["ci95"][1] * mg, 4)
out["growth_share_point"] = round(sl["b"] * mg / mw, 4) if mw != 0 else None
out["growth_share_ci"] = [round(sl["ci95"][0] * mg / mw, 4), round(sl["ci95"][1] * mg / mw, 4)] if mw != 0 else None
out["mde_80pct_5pct"] = round(2.8016 * sl["se"], 4)
return out
PERIODS = [(1990, 2007), (2007, 2022), (1997, 2007), (2007, 2017), (2017, 2022)]
GROUPS = {}
for k_i, (vname, mem) in enumerate(VARIANTS.items()):
GROUPS[vname] = {}
for p_i, (a, z) in enumerate(PERIODS):
GROUPS[vname][f"{a}-{z}"] = group_block(long_diff(a, z, s), mem, a, z, seed=20261029 + 100 * k_i + 10 * p_i)
R3["groups"] = GROUPS
# --- change 6: formal weighted-vs-unweighted test (DuMouchel-Duncan, HC3) and pairs bootstrap of the difference
def wdiff(d, members, seed, B=4999):
hi = keep_high(d, members)
hi = hi[hi["terc"] == "high"].reset_index(drop=True)
hi["w0s"] = hi["w0"] / hi["w0"].mean()
hi["gdp_w"] = hi["gdp"] * hi["w0s"]
dd = smf.ols("wedge ~ gdp + w0s + gdp_w", hi).fit(cov_type="HC3")
def slopes(ix):
y, x, w = hi["wedge"].values[ix], hi["gdp"].values[ix], hi["w0"].values[ix]
bu = np.polyfit(x, y, 1)[0]
bw = np.polyfit(x, y, 1, w=np.sqrt(w))[0]
return bw - bu
rng = np.random.default_rng(seed)
n = len(hi)
draws = np.array([slopes(rng.integers(0, n, n)) for _ in range(B)])
diff0 = float(slopes(np.arange(n)))
return {"diff_weighted_minus_unweighted": round(diff0, 4), "boot_se": round(float(draws.std(ddof=1)), 4),
"boot_ci95_pct": [round(float(np.quantile(draws, 0.025)), 4), round(float(np.quantile(draws, 0.975)), 4)],
"dumouchel_duncan_gdp_x_w": coef(dd, "gdp_w") | {"se_type": "HC3"}, "n": int(n)}
R3["weighted_vs_unweighted"] = {f"{a}-{z}": wdiff(long_diff(a, z, s), VARIANTS["all"], seed=20261129 + i)
for i, (a, z) in enumerate([(1990, 2007), (2007, 2022)])}
# --- change 3: group x period intercepts in annual FD; group x period FE (and country trends) in levels
fd3 = fd_frame(s)
for t_ in ["mid", "high"]:
for p_ in [0, 1]:
fd3[f"g_{t_}_{p_}"] = ((fd3["terc"] == t_) & (fd3["post"] == p_)).astype(float)
GPI = " + g_mid_0 + g_mid_1 + g_high_0 + g_high_1"
FD3 = {"n_obs": int(len(fd3))}
for y in ["co2", "consumption_co2", "wedge"]:
fml = f"d_{y} ~ C(year) + C(terc):C(post):d_gdp" + GPI
mdk = smf.ols(fml, fd3).fit(cov_type="hac-groupsum", cov_kwds={"time": fd3["tcode"].values, "maxlags": 2})
mcl = smf.ols(fml, fd3).fit(cov_type="cluster", cov_kwds={"groups": pd.factorize(fd3["country"])[0]})
FD3[y] = {"r2": round(float(mdk.rsquared), 4)}
for t_ in ["low", "mid", "high"]:
for p_ in [0, 1]:
nm = f"C(terc)[{t_}]:C(post)[{p_}]:d_gdp"
FD3[y][f"{t_}_{'post' if p_ else 'pre'}"] = {"dk": coef(mdk, nm), "cl_se": round(float(mcl.bse[nm]), 4)}
if y == "wedge":
FD3["wedge_high_intercepts_dk"] = {k: coef(mdk, k) for k in ["g_high_0", "g_high_1"]}
R3["fd_group_period_intercepts"] = FD3
lv3 = s.dropna(subset=["terc"]).copy()
lv3["post"] = (lv3["year"] >= 2008).astype(int)
lv3["g_mid_post"] = ((lv3["terc"] == "mid") & (lv3["post"] == 1)).astype(float)
lv3["g_high_post"] = ((lv3["terc"] == "high") & (lv3["post"] == 1)).astype(float)
lv3["trend"] = lv3["year"] - 1990
LV3 = {"n_obs": int(len(lv3))}
for lab, extra in [("group_period_fe", " + g_mid_post + g_high_post"), ("country_trends", " + C(country):trend")]:
mm = smf.ols("l_wedge ~ C(country) + C(year) + C(terc):C(post):l_gdp" + extra, lv3).fit(
cov_type="cluster", cov_kwds={"groups": pd.factorize(lv3["country"])[0]})
LV3[lab] = {"pre": coef(mm, "C(terc)[high]:C(post)[0]:l_gdp"), "post": coef(mm, "C(terc)[high]:C(post)[1]:l_gdp")}
R3["levels_fe"] = LV3
# --- change 5: does the file itself show the stated benchmark years? Kinks and dispersion of annual wedge changes
kk = s.sort_values(["country", "year"]).copy()
gk = kk.groupby("country")
ok = gk["year"].diff() == 1
kk["dw"] = gk["l_wedge"].diff().where(ok)
kk["dw_next"] = kk.groupby("country")["dw"].shift(-1).where(kk.groupby("country")["year"].shift(-1) == kk["year"] + 1)
kk["kink"] = (kk["dw_next"] - kk["dw"]).abs()
ys = kk.groupby("year").agg(kink=("kink", "median"), sd_dw=("dw", "std"))
ys = ys.loc[1992:2021].dropna()
bench_in = [y for y in BENCH if y in ys.index]
def perm(stat, seed, R=20000):
obs = float(ys.loc[bench_in, stat].mean())
rng = np.random.default_rng(seed)
vals = ys[stat].values
sims = np.array([vals[rng.choice(len(vals), len(bench_in), replace=False)].mean() for _ in range(R)])
rank = ys[stat].rank(ascending=False)
return {"bench_mean": round(obs, 4), "other_mean": round(float(ys.drop(bench_in)[stat].mean()), 4),
"perm_p_greater": round(float(np.mean(sims >= obs)), 4),
"bench_ranks_of_n": {str(y): int(rank[y]) for y in bench_in}, "n_years": int(len(ys))}
R3["benchmark_signature"] = {"kink_median_abs_second_diff": perm("kink", 20261229),
"dispersion_sd_dw": perm("sd_dw", 20261230),
"top5_kink_years": [int(y) for y in ys["kink"].sort_values(ascending=False).index[:5]],
"top5_sd_years": [int(y) for y in ys["sd_dw"].sort_values(ascending=False).index[:5]]}
# --- change 7: China's net embodied trade in the file
cn = df[(df["country"] == "China") & df["year"].between(1990, 2022)].set_index("year")["trade_co2"]
R3["china_trade_co2"] = {"y1990": round(float(cn.loc[1990]), 1), "y2005": round(float(cn.loc[2005]), 1),
"y2007": round(float(cn.loc[2007]), 1), "y2022": round(float(cn.loc[2022]), 1),
"years_negative": int((cn < 0).sum()), "years_observed": int(cn.notna().sum())}
REVISION["round3"] = R3
# --- round 4: how much of the top group's production emissions each composition variant removes
def co2_shares(y):
h = s[(s["year"] == y) & s["country"].isin(highs)].set_index("country")["co2"]
tot = float(h.sum())
pct = lambda names: round(100 * float(h[h.index.isin(names)].sum()) / tot, 1)
return {"n_members": int(len(h)), "total_mt": round(tot, 1), "oil_transition_pct": pct(OIL + TRANS),
"oil_pct": pct(OIL), "transition_pct": pct(TRANS), "united_states_pct": pct(["United States"]),
"australia_pct": pct(["Australia"]), "dropped_by_net_importers_1997_pct": pct([c_ for c_ in highs if c_ not in VARIANTS["net_importers_1997"]])}
REVISION["round4"] = {"high_group_co2_shares": {str(y): co2_shares(y) for y in [1990, 1997, 2007, 2017, 2022]},
"note": "shares of the 1990 top tercile's summed territorial co2 in the analysis sample (Kuwait 1990-1992 rows excluded)"}
print("round4 co2 shares:", REVISION["round4"]["high_group_co2_shares"])
print("round3 groups (LD wedge HC1 / HC3 / wild p):", {k: {p: (v[p]["slope_wedge_hc1"]["b"], v[p]["slope_wedge_hc1"]["se"], v[p]["slope_wedge_hc3"]["se_hc3"], v[p]["slope_wedge_wild_p"]) for p in ["1990-2007", "2007-2022"]} for k, v in GROUPS.items()})
print("round3 w-vs-u:", R3["weighted_vs_unweighted"])
print("round3 FD/levels high wedge:", FD3["wedge"]["high_pre"], FD3["wedge"]["high_post"], LV3)
print("round3 benchmark signature:", R3["benchmark_signature"])
# ================================================================ REVISION (referee round 5)
# Everything above is unchanged, so every earlier number reproduces. New randomness uses its own seeded generators.
R5 = {}
# --- change 3: can the file's own columns define the oil-exporter set? The file has no production or export column;
# flaring_co2 (CO2 from gas flared at oil and gas extraction sites) is its only extraction-specific column.
fl = df[df["country"].isin(highs) & df["year"].between(1990, 2022)].copy()
fl["flaring_share_pct"] = 100 * fl["flaring_co2"] / fl["co2"]
FLV = {"flaring_share_mean_1990_2022": fl.groupby("country")["flaring_share_pct"].mean(),
"flaring_per_capita_mean_1990_2022": fl.groupby("country")["flaring_co2_per_capita"].mean()}
for y in [1990, 1997, 2007]:
FLV[f"flaring_share_{y}"] = fl[fl["year"] == y].set_index("country")["flaring_share_pct"]
RULES = {}
for k, v in FLV.items():
v = v.dropna()
thr = float(v.reindex(OIL).min()) # lowest threshold that keeps all five named exporters
sel = sorted(v[v >= thr].index.tolist())
RULES[k] = {"threshold_min_over_oil_set": round(thr, 4), "selected": sel, "n_selected": len(sel),
"equals_oil_set": sorted(sel) == sorted(OIL), "includes_norway": "Norway" in sel,
"extra_beyond_oil_set": [c_ for c_ in sel if c_ not in OIL],
"norway_value": round(float(v["Norway"]), 4), "norway_rank_of_n": [int(v.rank(ascending=False)["Norway"]), int(len(v))]}
for y in [1997, 2007]:
tt = s[(s["year"] == y) & (s["terc"] == "high")].set_index("country")["trade_co2"]
sel = sorted(tt[tt <= 0].index.tolist())
RULES[f"net_exporter_embodied_{y}"] = {"selected": sel, "n_selected": len(sel), "equals_oil_set": sorted(sel) == sorted(OIL),
"includes_norway": "Norway" in sel, "oil_set_missed": [c_ for c_ in OIL if c_ not in sel],
"extra_beyond_oil_set": [c_ for c_ in sel if c_ not in OIL]}
R5["oil_rule_from_file"] = {"rules_tried": RULES, "any_rule_reproduces_oil_set": any(r_["equals_oil_set"] for r_ in RULES.values()),
"every_flaring_rule_includes_norway": all(RULES[k]["includes_norway"] for k in FLV),
"norway_first_year_in_sample": int(s.loc[s["country"] == "Norway", "year"].min()),
"note": "flaring rules: lowest threshold that keeps all five named oil exporters, over the 40 top-tercile countries, all file rows 1990-2022"}
# Core group that also drops Norway: same quantities and the same bootstrap seeds as the round-3 core rows, every period
NO_NOR = [c_ for c_ in VARIANTS["no_oil_transition"] if c_ != "Norway"]
R5["core_no_norway"] = {f"{a}-{z}": group_block(long_diff(a, z, s), NO_NOR, a, z, seed=20261029 + 100 * 1 + 10 * p_i) for p_i, (a, z) in enumerate(PERIODS)}
R5["core_no_norway_members"] = NO_NOR
def share_of(names, y):
h = s[(s["year"] == y) & s["country"].isin(highs)].set_index("country")["co2"]
return round(100 * float(h[h.index.isin(names)].sum()) / float(h.sum()), 1)
R5["norway_co2_share_pct"] = {str(y): share_of(["Norway"], y) for y in [2007, 2017, 2022]}
R5["oil_transition_norway_co2_share_pct"] = {str(y): share_of(OIL + TRANS + ["Norway"], y) for y in [1990, 1997, 2007, 2017, 2022]}
# --- change 5: net importers by the 1990 sign of trade_co2 (Kuwait 1990 and Norway lack 1990 rows in the sample)
t90 = s[(s["year"] == 1990) & (s["terc"] == "high")].set_index("country")["trade_co2"]
NI90 = sorted(t90[t90 > 0].index.tolist())
R5["net_importers_1990"] = {"members": NI90, "n": len(NI90), "n_with_1990_row": int(len(t90)),
"net_exporters_1990": sorted(t90[t90 <= 0].index.tolist()),
"in_1990_not_1997_set": [c_ for c_ in NI90 if c_ not in VARIANTS["net_importers_1997"]],
"in_1997_not_1990_set": [c_ for c_ in VARIANTS["net_importers_1997"] if c_ not in NI90],
"dropped_co2_share_pct": {str(y): share_of([c_ for c_ in highs if c_ not in NI90], y) for y in [1990, 2007]},
"groups": {f"{a}-{z}": group_block(long_diff(a, z, s), NI90, a, z, seed=20270201 + 10 * p_i)
for p_i, (a, z) in enumerate([(1990, 2007), (1997, 2007)])}}
# --- change 2: one inference rule (wild-bootstrap p, 10 percent) applied to every top-group slope, both directions
def rule_row(r_):
u, w_ = r_["slope_wedge_hc1"], r_["slope_wedge_weighted_hc1"]
return {"b": u["b"], "p_hc1": u["p"], "p_hc3": r_["slope_wedge_hc3"]["p_hc3"], "p_wild": r_["slope_wedge_wild_p"],
"sig10_hc1": u["p"] < 0.1, "sig10_wild": r_["slope_wedge_wild_p"] < 0.1,
"w_b": w_["b"], "w_p_hc1": w_["p"], "w_p_wild": r_["slope_wedge_weighted_wild_p"],
"w_sig10_hc1": w_["p"] < 0.1, "w_sig10_wild": r_["slope_wedge_weighted_wild_p"] < 0.1}
ALLG = dict(GROUPS) | {"core_no_norway": R5["core_no_norway"], "net_importers_1990": R5["net_importers_1990"]["groups"]}
INF = {k: {p: rule_row(v) for p, v in g_.items()} for k, g_ in ALLG.items()}
R5["inference_rule_wild_10pct"] = {"rows": INF,
"sig_under_wild": sorted(f"{k}|{p}" for k, g_ in INF.items() for p, r_ in g_.items() if r_["sig10_wild"]),
"sig_under_wild_weighted": sorted(f"{k}|{p}" for k, g_ in INF.items() for p, r_ in g_.items() if r_["w_sig10_wild"]),
"status_differs_hc1_vs_wild": sorted(f"{k}|{p}" for k, g_ in INF.items() for p, r_ in g_.items() if r_["sig10_hc1"] != r_["sig10_wild"]),
"weighted_status_differs_hc1_vs_wild": sorted(f"{k}|{p}" for k, g_ in INF.items() for p, r_ in g_.items() if r_["w_sig10_hc1"] != r_["w_sig10_wild"])}
REVISION["round5"] = R5
print("round5 oil rules:", {k: (v["n_selected"], v["includes_norway"], v["equals_oil_set"]) for k, v in RULES.items()})
print("round5 core no Norway:", {p: (v["n_high"], v["slope_wedge_hc1"]["b"], v["slope_wedge_hc1"]["se"], v["slope_wedge_wild_p"]) for p, v in R5["core_no_norway"].items()})
print("round5 net importers 1990:", NI90, {p: (v["n_high"], v["slope_wedge_hc1"]["b"], v["slope_wedge_hc1"]["se"], v["slope_wedge_wild_p"]) for p, v in R5["net_importers_1990"]["groups"].items()})
print("round5 wild-rule significant:", R5["inference_rule_wild_10pct"]["sig_under_wild"], R5["inference_rule_wild_10pct"]["sig_under_wild_weighted"])
# ================================================================ REVISION (referee round 6)
# Everything above is unchanged, so every earlier number reproduces. New randomness uses its own seeded generators.
R6 = {}
INF5 = R5["inference_rule_wild_10pct"]["rows"]
P_IDX = {f"{a}-{z}": i for i, (a, z) in enumerate(PERIODS)}
def slope_spec(key):
"""Members, endpoints and the bootstrap seed that produced the round-3/round-5 row named 'group|period'."""
g_, p = key.split("|")
a, z = map(int, p.split("-"))
if g_ in VARIANTS:
return VARIANTS[g_], a, z, 20261029 + 100 * list(VARIANTS).index(g_) + 10 * P_IDX[p]
if g_ == "core_no_norway":
return NO_NOR, a, z, 20261029 + 100 * 1 + 10 * P_IDX[p]
return NI90, a, z, 20270201 + 10 * [(1990, 2007), (1997, 2007)].index((a, z))
def high_frame(members, a, z):
dv = keep_high(long_diff(a, z, s), members)
return dv, dv[dv["terc"] == "high"].reset_index(drop=True)
def ols_diag(y, x, names):
"""OLS of y on x within the top group: leverage, studentized residuals, Cook's D, and each country's share of
the heteroskedasticity-robust variance numerator sum((x - xbar)^2 e^2)."""
y, x = np.asarray(y, float), np.asarray(x, float)
X = np.column_stack([np.ones_like(x), x])
XtXi = np.linalg.inv(X.T @ X)
e = y - X @ (XtXi @ X.T @ y)
h = np.einsum("ij,jk,ik->i", X, XtXi, X)
s2 = float(e @ e) / (len(y) - 2)
r = e / np.sqrt(s2 * (1 - h))
cook = r ** 2 / 2 * h / (1 - h)
v = (x - x.mean()) ** 2 * e ** 2
v = v / v.sum()
def top(a_):
i = int(np.argmax(a_))
return {"country": str(names[i]), "value": round(float(a_[i]), 4)}
return {"max_leverage": top(h) | {"mean_leverage": round(2 / len(y), 4)}, "max_abs_studentized_resid": top(np.abs(r)),
"max_cooks_d": top(cook), "max_share_hc_variance": top(v),
"per_country": {str(c_): {"dlog_gdp": round(float(x[i]), 4), "dw": round(float(y[i]), 4), "leverage": round(float(h[i]), 4),
"studentized_resid": round(float(r[i]), 4), "cooks_d": round(float(cook[i]), 4),
"share_hc_variance": round(float(v[i]), 4)} for i, c_ in enumerate(names)}}
def loo_block(key):
members, a, z, seed = slope_spec(key)
dv, hi = high_frame(members, a, z)
g_, p = key.split("|")
p_full = wild_p(hi["wedge"], hi["gdp"], None, seed)
assert p_full == INF5[g_][p]["p_wild"] and high_slope(dv)["b"] == INF5[g_][p]["b"]
rows = {}
for c_ in hi["country"]:
h_, c1 = hi[hi["country"] != c_], high_slope(dv[dv["country"] != c_])
rows[str(c_)] = {"b": c1["b"], "p_hc1": c1["p"],
"p_hc3": hc3_slope(h_["wedge"], h_["gdp"])["p_hc3"], "p_wild": wild_p(h_["wedge"], h_["gdp"], None, seed)}
lo = pd.DataFrame(rows).T
b0 = INF5[g_][p]["b"]
infl = str((lo["b"] - b0).abs().idxmax())
out = {"n": int(len(hi)), "b_full": b0, "p_wild_full": p_full, "seed": int(seed),
"loo_min": round(float(lo["b"].min()), 4), "loo_min_dropped": str(lo["b"].idxmin()),
"loo_max": round(float(lo["b"].max()), 4), "loo_max_dropped": str(lo["b"].idxmax()),
"most_influential": infl, "b_without_most_influential": float(lo.loc[infl, "b"]),
"p_hc1_without_most_influential": float(lo.loc[infl, "p_hc1"]), "p_hc3_without_most_influential": float(lo.loc[infl, "p_hc3"]),
"p_wild_without_most_influential": float(lo.loc[infl, "p_wild"]),
"loo_p_wild_range": [round(float(lo["p_wild"].min()), 4), round(float(lo["p_wild"].max()), 4)],
"n_loo_fits_sig10_wild": int((lo["p_wild"] < 0.1).sum()), "n_loo_fits": int(len(lo)),
"diagnostics": ols_diag(hi["wedge"], hi["gdp"], hi["country"].tolist()),
"loo": {c_: {k: float(v) for k, v in r_.items()} for c_, r_ in rows.items()}}
out["malta_in_sample"] = "Malta" in rows
if "Malta" in rows:
out["without_malta"] = {k: float(v) for k, v in rows["Malta"].items()}
return out
PASS = R5["inference_rule_wild_10pct"]["sig_under_wild"]
R6["leave_one_out_passing"] = {k: loo_block(k) for k in PASS}
R6["malta_drives_core_2007_2017"] = {k: R6["leave_one_out_passing"][k]["most_influential"] == "Malta"
for k in ["no_oil_transition|2007-2017", "core_no_norway|2007-2017"]}
# --- change 3: wild-bootstrap p for every top-group long-difference wedge slope outside Table 4 (Table 6 endpoints, Table 5 rows)
def rule_slope(hi, x, b_hc1, seed, weighted=None):
w_ = hi["w0"] if weighted else None
r_ = {"b": b_hc1["b"], "se_hc1": b_hc1["se"], "p_hc1": b_hc1["p"], "p_hc3": hc3_slope(hi["wedge"], hi[x], w_)["p_hc3"],
"p_wild": wild_p(hi["wedge"], hi[x], w_, seed)}
r_["sig10_hc1"], r_["sig10_wild"] = r_["p_hc1"] < 0.1, r_["p_wild"] < 0.1
return r_
EPF = {"avg3_1990-92_to_2005-07": long_diff_avg((1990, 1992), (2005, 2007), s), "benchmark_1997-2007": long_diff(1997, 2007, s),
"avg3_2005-07_to_2020-22": long_diff_avg((2005, 2007), (2020, 2022), s), "benchmark_2007-2017": long_diff(2007, 2017, s),
"end2019_2007-2019": long_diff(2007, 2019, s)}
EP_SEED = {"benchmark_1997-2007": 20261029 + 10 * P_IDX["1997-2007"], "benchmark_2007-2017": 20261029 + 10 * P_IDX["2007-2017"]}
T6 = {}
for i, (k, d) in enumerate(EPF.items()):
hi = d[d["terc"] == "high"]
sd = EP_SEED.get(k, 20270301 + 10 * i)
T6[k] = {"wedge": rule_slope(hi, "gdp", ENDPOINTS[k]["wedge"], sd),
"wedge_weighted": rule_slope(hi, "gdp", ENDPOINTS[k]["wedge_weighted"], sd + 1, weighted=True), "seed": int(sd)}
for k, g_p in [("benchmark_1997-2007", "1997-2007"), ("benchmark_2007-2017", "2007-2017")]:
assert T6[k]["wedge"]["p_wild"] == INF5["all"][g_p]["p_wild"] and T6[k]["wedge_weighted"]["p_wild"] == INF5["all"][g_p]["w_p_wild"]
R6["table6_endpoints_wild"] = T6
T5F = {"start1995_pre": (p95, "gdp", ROB["start1995_high_wedge_pre"]),
"pop1m_pre": (pre[pre["pop90"] >= 1e6], "gdp", ROB["pop1m_high_wedge_pre"]),
"pop1m_post": (post[post["pop90"] >= 1e6], "gdp", ROB["pop1m_high_wedge_post"]),
"percapita_pre": (pre, "gdppc", ROB["percapita_high_wedge_pre"]),
"percapita_post": (post, "gdppc", ROB["percapita_high_wedge_post"]),
"kuwait_keep_all_pre": (long_diff(1990, 2007, make_sample("keep_all")), "gdp", KEY["keep_all"]["ld_high_wedge"]["pre"]),
"kuwait_drop1991_92_pre": (long_diff(1990, 2007, make_sample("drop1991_92")), "gdp", KEY["drop1991_92"]["ld_high_wedge"]["pre"])}
T5 = {}
for i, (k, (d, x, c)) in enumerate(T5F.items()):
sd = 20270401 + 10 * (list(T5F).index("kuwait_keep_all_pre") if k == "kuwait_drop1991_92_pre" else i) # same rows, same seed
T5[k] = rule_slope(d[d["terc"] == "high"], x, c, sd) | {"n_high": int((d["terc"] == "high").sum())}
R6["table5_long_difference_wild"] = T5
# --- change 4: multiple testing for the wild-bootstrap rule
from scipy.stats import binom
def family(rows, alpha=0.10):
"""rows: list of (label, p). Expected passes under the global null, Bonferroni and Holm (step-down)."""
n = len(rows)
ps = np.sort(np.array([p_ for _, p_ in rows], float))
k_obs = int((ps < alpha).sum())
def holm(a_):
k_ = 0
for j, p_ in enumerate(ps):
if p_ <= a_ / (n - j):
k_ += 1
else:
break
return k_
holm_adj = np.maximum.accumulate(np.minimum(1.0, ps * (n - np.arange(n))))
return {"n_tests": n, "n_pass_10pct": k_obs, "expected_pass_under_null_10pct": round(alpha * n, 2),
"prob_ge_observed_if_independent": round(float(binom.sf(k_obs - 1, n, alpha)), 4),
"min_p": round(float(ps[0]), 4), "bonferroni_threshold_10pct": round(0.10 / n, 4), "bonferroni_threshold_5pct": round(0.05 / n, 4),
"bonferroni_min_adjusted_p": round(float(min(1.0, ps[0] * n)), 4), "holm_min_adjusted_p": round(float(holm_adj[0]), 4),
"n_reject_bonferroni_10pct": int((ps <= 0.10 / n).sum()), "n_reject_holm_10pct": holm(0.10), "n_reject_holm_5pct": holm(0.05),
"labels": [l_ for l_, _ in rows]}
def distinct(field):
seen, rows = set(), []
for g_, gr in INF5.items():
for p, r_ in gr.items():
sig = (p, r_["b" if field == "p_wild" else "w_b"], r_[field])
if sig not in seen:
seen.add(sig)
rows.append((f"{g_}|{p}", r_[field]))
return rows
FAM_T4 = distinct("p_wild")
FAM_T4W = distinct("w_p_wild")
FAM_ALL = FAM_T4 + [(f"table6|{k}", v["wedge"]["p_wild"]) for k, v in T6.items() if not k.startswith("benchmark")] + \
[(f"table5|{k}", v["p_wild"]) for k, v in T5.items() if k != "kuwait_drop1991_92_pre"]
R6["multiple_testing"] = {"table4_unweighted": family(FAM_T4), "table4_weighted": family(FAM_T4W),
"all_long_difference_unweighted": family(FAM_ALL),
"note": "table4: distinct rows of round5 inference_rule_wild_10pct (core_no_norway 1990-2007 and 1997-2007 equal the core rows); "
"all_long_difference adds table 6 avg3/end2019 rows and table 5 long-difference rows (Kuwait 1991-92 variant equals keep_all); "
"tests are nested and overlapping, not independent"}
assert R6["multiple_testing"]["table4_unweighted"]["n_tests"] == 20
# --- change 2: weighted intercepts (drift at the weighted mean growth) of the net-importer groups, reused from rounds 3 and 5
R6["weighted_intercepts_net_importers"] = {"net_importers_1997": {p: GROUPS["net_importers_1997"][p]["weighted_intercept_demeaned"] for p in ["1990-2007", "2007-2017"]},
"net_importers_1990": {p: R5["net_importers_1990"]["groups"][p]["weighted_intercept_demeaned"] for p in ["1990-2007", "1997-2007"]}}
REVISION["round6"] = R6
print("round6 LOO:", {k: (v["b_full"], v["loo_min"], v["loo_min_dropped"], v["loo_max"], v["loo_max_dropped"], v["most_influential"],
v["b_without_most_influential"], v["p_wild_without_most_influential"]) for k, v in R6["leave_one_out_passing"].items()})
print("round6 table6 wild p:", {k: (v["wedge"]["b"], v["wedge"]["p_wild"], v["wedge_weighted"]["p_wild"]) for k, v in T6.items()})
print("round6 table5 wild p:", {k: (v["b"], v["p_wild"]) for k, v in T5.items()})
print("round6 multiple testing:", {k: (v["n_tests"], v["n_pass_10pct"], v["expected_pass_under_null_10pct"], v["min_p"], v["bonferroni_threshold_10pct"], v["n_reject_holm_10pct"])
for k, v in R6["multiple_testing"].items() if k != "note"})
# ================================================================ REVISION (referee round 7)
# Everything above is unchanged, so every earlier number reproduces. New randomness uses its own seeded generators.
# Malta deletion for every top-group long-difference wedge slope (Tables 3-6), unweighted and weighted, with the
# bootstrap seed of the full-sample row; Malta's influence in each sample; the two-period comparison and the
# aggregate wedge changes without Malta; and a symmetric Cook's-distance screen (sensitivity check only).
R7 = {}
DROP = "Malta"
mt = df[(df["country"] == DROP) & df["year"].between(1990, 2022)].set_index("year")
R7["malta_facts"] = {str(y): {"co2_mt": round(float(mt.loc[y, "co2"]), 3), "consumption_co2_mt": round(float(mt.loc[y, "consumption_co2"]), 3),
"wedge": round(float(np.log(mt.loc[y, "consumption_co2"] / mt.loc[y, "co2"])), 4),
"population": int(mt.loc[y, "population"])} for y in [1990, 1997, 2007, 2017, 2019, 2022]}
R7["malta_facts"]["tercile"] = str(terc.get(DROP))
def diag_w(y, x, w=None):
"""OLS (or WLS with weights w) of y on x: leverage, studentized residual, Cook's D and each observation's share of
the slope's robust variance numerator sum(w^2 (x - xbar_w)^2 e^2); with w None these equal ols_diag's measures."""
y, x = np.asarray(y, float), np.asarray(x, float)
sw = np.sqrt(np.ones_like(y) if w is None else np.asarray(w, float))
X, Y = np.column_stack([np.ones_like(x), x]) * sw[:, None], y * sw
XtXi = np.linalg.inv(X.T @ X)
e = Y - X @ (XtXi @ X.T @ Y)
h = np.einsum("ij,jk,ik->i", X, XtXi, X)
r = e / np.sqrt(float(e @ e) / (len(y) - 2) * (1 - h))
v = ((XtXi @ X.T)[1] * e) ** 2
return h, r, r ** 2 / 2 * h / (1 - h), v / v.sum()
def with_country(d):
return d if "country" in d.columns else d.reset_index().rename(columns={"index": "country"})
def fit_row(d, x, seed, weighted):
hi = d[d["terc"] == "high"]
w_ = hi["w0"] if weighted else None
c = high_wls(d) if weighted else high_slope(d, x=x)
h3 = hc3_slope(hi["wedge"], hi[x], w_)
return {"n": int(len(hi)), "b": c["b"], "se_hc1": c["se"], "p_hc1": c["p"], "se_hc3": h3["se_hc3"], "p_hc3": h3["p_hc3"],
"p_wild": wild_p(hi["wedge"], hi[x], w_, seed)}
def malta_row(d, x, seed, weighted=False):
"""Full-sample and Malta-deleted slope, Malta's influence in the full sample, and the symmetric screen."""
d = with_country(d)
hi = d[d["terc"] == "high"].reset_index(drop=True)
names = hi["country"].tolist()
h, r, cook, v = diag_w(hi["wedge"], hi[x], hi["w0"] if weighted else None)
out = {"seed": int(seed), "full": fit_row(d, x, seed, weighted), "malta_in_sample": DROP in names}
n = len(names)
cut = 4 / n
drop = [names[i] for i in range(n) if cook[i] > cut]
scr = fit_row(d[~d["country"].isin(drop)], x, seed, weighted)
out["screen_cooks_d_4_over_n"] = {"cutoff": round(cut, 4), "dropped": drop} | {k: scr[k] for k in ["n", "b", "p_hc3", "p_wild"]}
i_max = int(np.argmax(v))
out["largest_share_hc_variance"] = {"country": names[i_max], "value": round(float(v[i_max]), 4)}
if DROP in names:
i = names.index(DROP)
rank = lambda a_: int((a_ > a_[i]).sum() + 1)
out["without_malta"] = fit_row(d[d["country"] != DROP], x, seed, weighted)
out["malta"] = {"leverage": round(float(h[i]), 4), "mean_leverage": round(2 / n, 4), "studentized_resid": round(float(r[i]), 4),
"cooks_d": round(float(cook[i]), 4), "share_hc_variance": round(float(v[i]), 4),
"rank_share_hc_variance": rank(v), "rank_abs_studentized_resid": rank(np.abs(r)), "rank_cooks_d": rank(cook),
"dlog_gdp": round(float(hi.loc[i, x]), 4), "dw": round(float(hi.loc[i, "wedge"]), 4)}
if weighted:
out["malta"]["weight_share"] = round(float(hi.loc[i, "w0"] / hi["w0"].sum()), 5)
return out
# Table 4 (its all-group 1990-2007 and 2007-2022 rows are Table 3's top-group rows): every group and period, seeds as in rounds 3 and 5
T4 = {}
for g_, gr in INF5.items():
T4[g_] = {}
for p in gr:
members, a, z, sd = slope_spec(f"{g_}|{p}")
dv = keep_high(long_diff(a, z, s), members)
T4[g_][p] = {"unweighted": malta_row(dv, "gdp", sd), "weighted": malta_row(dv, "gdp", sd + 1, weighted=True)}
u, w_ = T4[g_][p]["unweighted"]["full"], T4[g_][p]["weighted"]["full"]
assert (u["b"], u["p_wild"], w_["b"], w_["p_wild"]) == (gr[p]["b"], gr[p]["p_wild"], gr[p]["w_b"], gr[p]["w_p_wild"])
for k in PASS: # the four passing slopes: Malta-deleted fits equal round 6's leave-one-out fits
g_, p = k.split("|")
wm, lo = T4[g_][p]["unweighted"]["without_malta"], R6["leave_one_out_passing"][k]["loo"]["Malta"]
assert (wm["b"], wm["p_hc3"], wm["p_wild"]) == (lo["b"], lo["p_hc3"], lo["p_wild"])
R7["table4"] = T4
# Table 6 endpoints and Table 5 long-difference variants, seeds as in round 6
R7["table6"] = {}
for i, (k, d) in enumerate(EPF.items()):
sd = T6[k]["seed"]
R7["table6"][k] = {"unweighted": malta_row(d, "gdp", sd), "weighted": malta_row(d, "gdp", sd + 1, weighted=True)}
assert R7["table6"][k]["unweighted"]["full"]["p_wild"] == T6[k]["wedge"]["p_wild"]
assert R7["table6"][k]["weighted"]["full"]["p_wild"] == T6[k]["wedge_weighted"]["p_wild"]
R7["table5"] = {}
for i, (k, (d, x, c)) in enumerate(T5F.items()):
sd = 20270401 + 10 * (list(T5F).index("kuwait_keep_all_pre") if k == "kuwait_drop1991_92_pre" else i)
R7["table5"][k] = malta_row(d, x, sd)
assert R7["table5"][k]["full"]["p_wild"] == T5[k]["p_wild"]
# Table 3: weighted minus unweighted slope (pairs bootstrap, DuMouchel-Duncan) without Malta, round 3 seeds
NO_MT = [c_ for c_ in highs if c_ != DROP]
R7["table3_weighted_minus_unweighted_without_malta"] = {f"{a}-{z}": wdiff(long_diff(a, z, s), NO_MT, seed=20261129 + i)
for i, (a, z) in enumerate([(1990, 2007), (2007, 2022)])}
# Two-period comparison in the four groups with both periods, with and without Malta (stacked test of equation 3)
def compare(members, p1, p2, g_):
out = {}
for lab, mem in [("full", members), ("without_malta", [c_ for c_ in members if c_ != DROP])]:
d1, d2 = keep_high(long_diff(*p1, s), mem), keep_high(long_diff(*p2, s), mem)
r1, r2 = (T4[g_][f"{a}-{z}"]["unweighted"][lab] for a, z in [p1, p2])
w1, w2 = (T4[g_][f"{a}-{z}"]["weighted"][lab] for a, z in [p1, p2])
out[lab] = {"first": r1["b"], "second": r2["b"], "second_minus_first": round(r2["b"] - r1["b"], 4), "second_exceeds_first": r2["b"] > r1["b"],
"stacked": stacked(d1, d2), "weighted_first": w1["b"], "weighted_second": w2["b"], "weighted_second_exceeds_first": w2["b"] > w1["b"]}
return out
CMP_G = {"all": VARIANTS["all"], "no_oil_transition": VARIANTS["no_oil_transition"], "core_no_norway": NO_NOR,
"net_importers_1997": VARIANTS["net_importers_1997"]}
R7["period_comparison"] = {f"{p1[0]}-{p1[1]}_vs_{p2[0]}-{p2[1]}": {g_: compare(m_, p1, p2, g_) for g_, m_ in CMP_G.items()}
for p1, p2 in [((1990, 2007), (2007, 2022)), ((1997, 2007), (2007, 2017))]}
assert R7["period_comparison"]["1990-2007_vs_2007-2022"]["all"]["full"]["stacked"] == LD["stacked"]
# Aggregate (summed) wedge changes without Malta, every group and period of Table 4 and the 2007-2019 row of Table 2
AGG7 = {}
for g_, gr in INF5.items():
AGG7[g_] = {}
for p in gr:
members, a, z, _ = slope_spec(f"{g_}|{p}")
AGG7[g_][p] = {"full": agg_wedge(a, z, s, members), "without_malta": agg_wedge(a, z, s, [c_ for c_ in members if c_ != DROP])}
AGG7["all"]["2007-2019"] = {"full": agg_wedge(2007, 2019, s, highs), "without_malta": agg_wedge(2007, 2019, s, NO_MT)}
R7["aggregate_wedge"] = AGG7
# Summary: which rows Malta dominates, and which slopes pass the wild-bootstrap rule with and without Malta
ROWS7 = [(f"table4|{g_}|{p}|{wt}", r_) for g_, gr in T4.items() for p, rr in gr.items() for wt, r_ in rr.items()] + \
[(f"table6|{k}|{wt}", r_) for k, rr in R7["table6"].items() for wt, r_ in rr.items()] + \
[(f"table5|{k}|unweighted", r_) for k, r_ in R7["table5"].items()]
MAL = [(l_, r_) for l_, r_ in ROWS7 if r_["malta_in_sample"]]
SEEN7, DIST7 = set(), [] # distinct samples: repeated rows (core without Norway before 2007, Table 6 benchmarks, Kuwait 1991-92) once
for l_, r_ in ROWS7:
sig = (l_.endswith("weighted") and not l_.endswith("unweighted"), r_["full"]["n"], r_["full"]["b"], r_["full"]["se_hc1"], r_["full"]["p_wild"])
if sig not in SEEN7:
SEEN7.add(sig)
DIST7.append((l_, r_))
R7["summary"] = {
"n_rows": len(ROWS7), "n_distinct_rows": len(DIST7), "n_rows_with_malta": len(MAL),
"rows_without_malta_in_sample": [l_ for l_, r_ in ROWS7 if not r_["malta_in_sample"]],
"distinct_malta_largest_share_hc_variance": sorted(l_ for l_, r_ in DIST7 if r_["malta_in_sample"] and r_["malta"]["rank_share_hc_variance"] == 1),
"distinct_pass_wild_10pct_full": sorted(l_ for l_, r_ in DIST7 if r_["full"]["p_wild"] < 0.1),
"distinct_pass_wild_10pct_after_screen": sorted(l_ for l_, r_ in DIST7 if r_["screen_cooks_d_4_over_n"]["p_wild"] < 0.1),
"malta_largest_share_hc_variance": sorted(l_ for l_, r_ in MAL if r_["malta"]["rank_share_hc_variance"] == 1),
"malta_share_hc_variance_over_half": sorted(l_ for l_, r_ in MAL if r_["malta"]["share_hc_variance"] > 0.5),
"malta_cooks_d_over_4_over_n": sorted(l_ for l_, r_ in MAL if r_["malta"]["cooks_d"] > 4 / r_["full"]["n"]),
"unweighted_malta_share_range": [min(r_["malta"]["share_hc_variance"] for l_, r_ in MAL if l_.endswith("|unweighted")),
max(r_["malta"]["share_hc_variance"] for l_, r_ in MAL if l_.endswith("|unweighted"))],
"weighted_malta_share_max": max(r_["malta"]["share_hc_variance"] for l_, r_ in MAL if l_.endswith("|weighted")),
"pass_wild_10pct_full": sorted(l_ for l_, r_ in ROWS7 if r_["full"]["p_wild"] < 0.1),
"pass_wild_10pct_without_malta": sorted(l_ for l_, r_ in MAL if r_["without_malta"]["p_wild"] < 0.1),
"pass_hc3_10pct_without_malta": sorted(l_ for l_, r_ in MAL if r_["without_malta"]["p_hc3"] < 0.1),
"pass_wild_10pct_after_screen": sorted(l_ for l_, r_ in ROWS7 if r_["screen_cooks_d_4_over_n"]["p_wild"] < 0.1)}
FAM7 = []
for l_ in R6["multiple_testing"]["table4_unweighted"]["labels"]:
g_, p = l_.split("|")
r_ = T4[g_][p]["unweighted"]
FAM7.append((l_, (r_["without_malta"] if r_["malta_in_sample"] else r_["full"])["p_wild"]))
R7["multiple_testing_table4_unweighted_without_malta"] = family(FAM7)
R7["note"] = ("Malta deletions use the seed of the full-sample row; diagnostics are Malta's in the full sample (weighted rows: WLS with "
"base-year co2 weights); the Cook's-distance screen drops every country with D > 4/n in each sample, once, and is a "
"sensitivity check only. All round-7 quantities were added after the main results were known.")
REVISION["round7"] = R7
print("round7 Malta deletion (b, p_hc3, p_wild full -> without Malta; Malta share):",
{l_: (r_["full"]["b"], r_["full"]["p_hc3"], r_["full"]["p_wild"], r_["without_malta"]["b"], r_["without_malta"]["p_hc3"],
r_["without_malta"]["p_wild"], r_["malta"]["share_hc_variance"]) for l_, r_ in MAL if l_.endswith("unweighted")})
print("round7 period comparison:", {k: {g_: (v["full"]["second_minus_first"], v["full"]["stacked"]["p"], v["without_malta"]["second_minus_first"],
v["without_malta"]["stacked"]["p"]) for g_, v in c_.items()} for k, c_ in R7["period_comparison"].items()})
print("round7 summary:", R7["summary"])
# ---------------------------------------------------------------- verdict
w_pre, w_post = LD["1990-2007"]["wedge"]["high"], LD["2007-2022"]["wedge"]["high"]
part1 = w_pre["b"] > 0 and w_pre["p"] < 0.05
part2 = w_post["b"] < w_pre["b"]
result = {
"sample": {"iso_entities": int(n_ent), "countries": int(s["country"].nunique()), "country_years": int(len(s)),
"kuwait_rows_dropped": int(kuw.sum()), "terciles_n": {k: int(v) for k, v in terc.value_counts().items()},
"tercile_cutoffs_gdppc_1990": [round(float(q1), 1), round(float(q2), 1)],
"accounting_max_abs_gap_mt": float(acc.max())},
"long_differences": LD, "annual_first_differences": FD, "trade_shares": shares, "robustness": ROB, "revision": REVISION,
"test": {"high_wedge_pre_positive_5pct": bool(part1), "high_wedge_post_smaller": bool(part2),
"high_wedge_change_p_stacked": LD["stacked"]["p"]},
"hypothesis_supported": bool(part1 and part2),
}
print("RESULT_JSON: " + json.dumps(result, separators=(",", ":")))
5.Analysis
Analysis complete: experiment succeeded; inconclusive; 22 key numbers.
Results and analysis
The run succeeded (119 countries, 3,905 country-years, 1990-2022). No version of the top group has a 1990-2007 wedge elasticity positive at 5 percent, so the refutation criterion is met, but displacement cannot be ruled out: the verdict is inconclusive. The full top tercile gives 0.027 (0.117), wild-bootstrap p = 0.81; the upper bound times mean growth (0.165) exceeds the mean drift (0.121), and the minimum detectable effect is 0.33. Post hoc variants give 0.305 (0.234, core group) and 0.105 (0.120, 1990 net importers). The larger 2007-2022 slopes are Malta's: it carries 64-87 percent of the robust variance term in every unweighted sample from 2007 including it. Without Malta, a post hoc deletion like that of the United Arab Emirates (UAE), the 2007-2022 slopes are -0.0075 to 0.158 and the 1990-2007 versus 2007-2022 ordering reverses in every group; between the 1997-2007 and 2007-2017 benchmarks it is mixed (full group 0.074 vs 0.125; core 0.348 vs 0.304; 1997 net importers 0.298 vs 0.144). No stacked test distinguishes the periods (p = 0.39-0.80 without Malta, 0.44-0.52 with). The Malta-deleted 2007-2017 slopes of the full group and the 1997 net importers are sensitive to the UAE: the Cook's-distance screen drops exactly Malta and the UAE there and gives 0.228 (wild p 0.037) and 0.344 (0.032). Four of 20 Table 4 slopes pass the wild-bootstrap 10 percent rule, from two pairs of overlapping groups with HC3 p of 0.21-0.31; about 2 are expected under the null and none survives Bonferroni or Holm. All four still pass after deleting their most influential country: without Malta the core 2007-2017 slopes fall to 0.304 and 0.339 (wild p 0.068 and 0.044); without the UAE the net-importer 1997-2007 slopes rise to 0.512 and 0.562 (wild p 0.069 and 0.036). For the net importers this says little: their passes survive only 15 of 27 and 5 of 28 single-country deletions, and the 1990-sign pass fails without Malta (wild p 0.104). Post hoc deletions are not evidence either way; where one country dominates, the wild-bootstrap p may be unreliable. Aggregate wedges rose 0.052-0.067 over 1990-2007 and fell 0.018-0.040 over 2007-2017.
Key numbers
| Metric | Value | Interpretation |
|---|---|---|
| Top-tercile wedge elasticity (all 40), long difference 1990-2007 | 0.0274 (HC1 0.1174; HC3 0.1375); HC1 95% CI [-0.2027, 0.2575]; wild-bootstrap p = 0.8086; 38 countries; MDE 0.3289 | Near zero and insignificant, but the upper bound times mean growth (0.258 x 0.640 = 0.165) exceeds the mean drift of 0.121, so growth could account for all of the drift. |
| Inference rule for top-group long-difference wedge slopes (Tables 3-6) | Wild-bootstrap p (Webb weights, null imposed, 9,999 draws) is the reference, at 10 percent, in both directions, for every long-difference wedge slope of the top group in Tables 3-6. Of the 20 distinct unweighted slopes in Table 4 (22 rows; the core without Norway repeats the core's 1990-2007 and 1997-2007 rows), four pass: core 2007-2017 0.9852 (HC1 0.6631, HC1 p = 0.1373, HC3 p = 0.2181, wild p = 0.0675); core without Norway 2007-2017 1.0243 (0.6708, HC3 p = 0.2092, wild p = 0.0514); 1997 net importers 1997-2007 0.2786 (0.1497, HC3 p = 0.2845, wild p = 0.0908); 1990 net importers 1997-2007 0.2935 (0.1595, HC3 p = 0.3125, wild p = 0.0971). No weighted slope has wild p < 0.10. | The rule strengthens two estimates that HC1 leaves insignificant and removes the HC1 stars from every weighted slope. The four come from two pairs of overlapping groups (core with and without Norway; net importers by the 1997 and 1990 signs, which differ by three countries). Intercepts, territorial and consumption slopes, annual and levels specifications are outside the rule. |
| Leave-one-out: core group 2007-2017 slope (29 countries; 28 without Norway) | Full sample 0.9852 (wild p = 0.0675). Leave-one-out range [0.3039 (Malta), 1.2654 (Greece)]; dropping any country other than Malta gives 0.870-1.265. Without Malta: 0.3039, HC1 p = 0.0686, HC3 p = 0.0912, wild p = 0.0684. Malta: Delta ln Y 0.4536, Delta w 1.5007, leverage 0.175 (mean 0.069), studentized residual 4.43, Cook's distance 2.08, 76.1% of the robust variance term sum (x - xbar)^2 e^2; largest leverage Greece (0.312). Wild p below 0.10 in 27 of 29 fits; not when Singapore (0.1425) or Portugal (0.1475) is dropped. Without Norway: 1.0243, range [0.3387 (Malta), 1.3156 (Singapore)]; without Malta 0.3387, HC3 p = 0.0656, wild p = 0.0436; Malta 74.4% of the variance term; 26 of 28 fits pass. | Malta alone drives the size of the slope; without it the slope is about the core group's 1990-2007 estimate of 0.305, and HC3 and the wild bootstrap then agree (HC3 p 0.091 and 0.066, wild p 0.068 and 0.044). With Malta in, the wild p (0.068) is far below the HC3 p (0.218); when one residual dominates, the wild-bootstrap p may be unreliable. Both slopes pass with and without Malta; what Malta drives is their size. The Malta deletion was chosen after the fact, like the UAE deletion for the net importers. |
| Leave-one-out: net-importer 1997-2007 slopes | 1997 sign (27): 0.2786, range [0.2025 (Singapore), 0.5116 (United Arab Emirates)]; UAE leverage 0.585 (Delta ln Y 1.2544), 44.2% of the variance term, slope without it 0.5116, HC3 p 0.0590, wild p 0.0687; largest studentized residual Singapore (2.83); leave-one-out wild p 0.0687-0.1444, below 0.10 in 15 of 27. 1990 sign (28): 0.2935, range [0.2225 (Singapore), 0.5617 (UAE)]; UAE leverage 0.574, 54.3%, slope without it 0.5617, HC3 p 0.0275, wild p 0.0360; Singapore residual 3.00; wild p 0.0360-0.1603, below 0.10 in 5 of 28. Without Malta: 0.2983 (wild p 0.0996) and 0.3127 (0.1035). | Without the United Arab Emirates, an oil exporter removed from the core group, the slopes rise to 0.512 and 0.562 (HC3 p 0.059 and 0.028; wild p 0.069 and 0.036). With it, its leverage (0.585, 0.574) and variance shares (44 and 54 percent) mean the full-sample wild p (0.091, 0.097; HC3 p 0.28, 0.31) may be unreliable. Both slopes pass with and without the UAE, but deleting it raises the slopes and lowers p, so surviving that deletion says little about robustness; without Malta the 1990-sign pass fails (wild p 0.1035). This deletion and Malta's were both chosen after the fact; neither the smaller nor the larger slopes are evidence for or against displacement. Passing the 10 percent rule is borderline: most single deletions push p just above 0.10. |
| Wild-bootstrap p for Table 6 endpoints and Table 5 long-difference variants | Table 6 wedge slope (unweighted / weighted): 1990-92 to 2005-07 0.882 / 0.618; 1997-2007 0.613 / 0.504; 2005-07 to 2020-22 0.876 / 0.354; 2007-2017 0.213 / 0.456; 2007-2019 0.272 / 0.601 (0.5355 has HC1 p = 0.0511, HC3 p = 0.0810). Table 5: first period from 1995 0.324; population at least 1 million 0.767 and 0.859; per-capita regressor 0.495 and 0.421; Kuwait 1990 kept 0.980. | Under the rule no long-difference wedge slope in Tables 5 or 6 is significant at 10 percent; the 2019-endpoint slope of 0.536 is significant only under HC1. |
| Multiple testing | Table 4 unweighted: 20 tests, 4 with wild p < 0.10, 2.0 expected if every null holds; probability of 4 or more if independent 0.133. Bonferroni threshold 0.10/20 = 0.005 (0.0025 at 5 percent); smallest wild p 0.0514; Bonferroni- and Holm-adjusted minimum p both 1; none significant after correction. Table 4 weighted: 20 tests, 0 pass, smallest p 0.1395. Adding the Table 5 and 6 long-difference slopes: 29 tests, 4 pass, 2.9 expected, probability 0.329, Bonferroni threshold 0.0034. With Malta deleted from every Table 4 sample: 3 of 20 pass, smallest p 0.0436, probability 0.3231 if independent, none after Bonferroni or Holm. | The tests are nested in time and overlap in sample, so they are not independent and the independent-case probability is only a reference; no slope is significant familywise. |
| Malta deletion for every top-group long-difference wedge slope (Tables 3-6) | Malta dominates all 11 distinct unweighted samples starting in 2007 that include it (Table 4, 2007-2022 and 2007-2017, for the four groups with both periods; Table 5 per-capita 2007-2022; Table 6 2005-07 to 2020-22 and 2007-2019; the sample restricted to populations of at least 1 million starts in 2007 but excludes Malta): share of the robust variance term 64.0-86.7%, the largest studentized residual (3.90-5.25), Cook's distance 0.79-2.25 against 4/n of 0.10-0.15; elsewhere its share is at most 3.2% (unweighted) and 2.4% (weighted). 2007-2022 without Malta: all -0.0075 (HC3 p 0.8961, wild p 0.8847), core 0.1018 (0.5305, 0.5048), core without Norway 0.1576 (0.2451, 0.2326), 1997 net importers 0.0208 (0.9189, 0.9176), against 0.1866, 0.7383, 0.7809, 0.5270 with Malta. 1990-2007 without Malta: 0.0259, 0.2974, 0.2974, 0.1282 (HC3 p 0.33-0.85, wild p 0.21-0.83). 2007-2017 without Malta: 0.1247, 0.3039, 0.3387, 0.1435. Tables 5-6 post-2007 without Malta: per-capita 0.0254, 2005-07 to 2020-22 -0.0906, 2007-2019 0.2856. Weighted slopes move by at most 0.0162 and none passes the rule either way. Malta 2007 to 2022: co2 2.728 to 1.774 Mt, consumption_co2 5.128 to 16.119 Mt. | Malta alone produces the post-2007 pattern: without it the 2007-2022 slopes are small and insignificant and the 1990-2007 slopes imprecise. The Malta deletion was chosen after the fact, on the same terms as the UAE deletion, and its 2007-2017 slopes for the full group and the 1997 net importers are sensitive to the UAE: in those two samples the Cook's-distance screen drops exactly Malta and the UAE and gives 0.2275 (wild p 0.0371) and 0.3441 (0.0320), while the core group's Malta-deleted 2007-2017 slopes (0.3039, 0.3387) still pass the rule. Post hoc deletions in either direction are not evidence. In every sample Malta dominates, the wild-bootstrap p may be unreliable; its size under one dominant residual was not examined. |
| Two-period comparison with and without Malta (stacked, equation 3, clustered by country) | 2007-2022 minus 1990-2007, with Malta (p) / without (SE; p): all 0.1592 (0.5201) / -0.0333 (0.1286; 0.7954); core 0.4328 (0.4840) / -0.1956 (0.2265; 0.3877); core without Norway 0.4754 (0.4407) / -0.1398 (0.2341; 0.5506); 1997 net importers 0.3967 (0.4367) / -0.1074 (0.1661; 0.5177). 2007-2017 minus 1997-2007 without Malta (1997-2007 vs 2007-2017 slopes; p): all 0.0740 vs 0.1247, 0.0507 (0.7592); core 0.3481 vs 0.3039, -0.0442 (0.9138); core without Norway 0.3481 vs 0.3387, -0.0095 (0.9819); 1997 net importers 0.2983 vs 0.1435, -0.1548 (0.3516); with Malta p 0.3185-0.5201. | Without Malta the 1990-2007 versus 2007-2022 ordering reverses in every group. On the benchmark endpoints it is mixed: not reversed in the full group, about equal in the core group with and without Norway, reversed for the 1997 net importers. No test distinguishes the periods with or without Malta. |
| Aggregate (summed) wedge changes without Malta | Every group and period within 0.0019 of its full-sample value; e.g. all 1990-2007 0.0637 (0.0638 with Malta), 2007-2017 -0.0216 (-0.0209), 1997 net importers 2007-2017 -0.0412 (-0.0398), 2007-2022 0.0421 (0.0440). | Descriptive totals; Malta's emissions are too small to move them. |
| Symmetric Cook's distance screen (drop D > 4/n in every sample; sensitivity check only) | 8 of 52 distinct top-group long-difference slopes (unweighted and weighted) pass the wild rule after the screen, against 4 in the full samples: four before 2007 (1997 net importers 1990-2007, unweighted and weighted; 1990 net importers 1997-2007, 0.3840, wild p 0.0828; first period from 1995) and four after (all 2007-2017, unweighted 0.2275, wild p 0.0371, and weighted; 1997 net importers 2007-2017, 0.3441, wild p 0.0320; in both 2007-2017 samples the screen drops exactly Malta and the United Arab Emirates; 2005-07 to 2020-22 weighted). Of the original four only the 1990 net importers' 1997-2007 slope still passes. | The screen reshuffles which slopes pass, before and after 2007 alike; it is a deletion rule too and drives no conclusion. It does show that the small Malta-deleted 2007-2017 slopes of the full group and the 1997 net importers depend on keeping the UAE; in the full group's 2007-2022 sample it drops Malta and Venezuela instead (-0.0158). |
| Core group (without the 5 oil exporters and 6 transition economies), defined after the main results were known | 1990-2007: 0.3055 (0.2337), CI [-0.1525, 0.7635], wild p = 0.3367, MDE 0.6547, 28 countries; 2007-2022: 0.7383 (0.5506), wild p = 0.2132, 29 countries | Too imprecise to distinguish from the full-group 0.027; the interval contains it. |
| Oil-exporter rule from the file's columns | No rule reproduces the five. Lowest flaring thresholds that keep all five select 8-20 countries, always including Norway: mean flaring share of co2 1990-2022 (threshold 1.36%) and 1997 share (1.63%) select 9 (adding Australia, Canada, New Zealand, Norway); mean flaring per person 1990-2022 (0.263 t) selects 8 (adding Australia, Canada, Norway); 1990 and 2007 shares select 20 and 15. Net-exporter status of embodied CO2 in 1997 or 2007 misses the United Arab Emirates and adds 8 or 3 others. Norway's mean flaring share, 3.94%, ranks 2nd of 40. | The oil-exporter set rests on an outside criterion (exports dominated by oil and gas), which Norway also meets; Norway's omission from the core group was an inconsistency. |
| Core group also without Norway (28 countries) | 1990-2007 and 1997-2007 identical to the core rows (Norway's consumption series begins in 2003). 2007-2022: 0.7809 (0.5423), HC3 0.6698, wild p = 0.1213, weighted -0.0377 (0.0820), Kish 3.98, aggregate 0.0398; 2007-2017: 1.0243 (0.6708), HC3 0.8158, wild p = 0.0514, weighted 0.0622 (0.1650), weighted intercept -0.0196 (0.0125), aggregate -0.0185; 2017-2022: -0.6090 (0.4050), wild p = 0.1571, aggregate 0.0582 | Dropping Norway changes nothing of substance; Norway emits 0.3 percent of the group's territorial CO2 in 2007, 2017 and 2022. |
| Net importers of embodied CO2 by the 1990 sign (28 of the 38 members with 1990 rows) | 1990-2007: 0.1049 (0.1197), CI [-0.1297, 0.3395], HC3 0.1301, wild p = 0.3221, weighted 0.1222 (0.0982), Kish 8.54, mean drift 0.1213 (0.0444), weighted intercept 0.0502 (0.0277), aggregate 0.0521; 1997-2007: 0.2935 (0.1595), HC3 p = 0.3125, wild p = 0.0971, aggregate 0.0635. Versus the 1997-sign group: adds Cyprus and Slovakia, drops Taiwan. | Selecting at the start of 1990-2007 instead of inside it gives a similar, insignificant slope and the same aggregate rise; the dropped members emit 52.9 percent of the group's CO2 in 1990. |
| Aggregate wedge change (balanced): all / core / core without Norway / 1997 net importers / 1990 net importers | 1990-2007: 0.0638 / 0.0666 / 0.0666 / 0.0539 / 0.0521; 2007-2017: -0.0209 / -0.0191 / -0.0185 / -0.0398 / not estimated; 2017-2022: 0.0530 / 0.0572 / 0.0582 / 0.0838 | Descriptive totals, consistent across group definitions; consistency is not precision. |
| Weighted intercept (drift at weighted mean growth), 2007-2017 | All -0.0181 (0.0117), p = 0.1206, Kish n 4.67; core -0.0202 (0.0126), p = 0.1092, Kish 4.01; core without Norway -0.0196 (0.0125), p = 0.1164, Kish 3.98; 1997 net importers -0.0429 (0.0157), p = 0.0062, Kish 10.14; net-importer groups, 1990-2007: 1990 sign 0.0502 (0.0277), p = 0.0700; 1997 sign 0.0594 (0.0282), p = 0.0351 | The full and core groups' 2007-2017 falls are not distinguishable from zero, before allowing for inventory error; only the 1997 net importers' fall is, under HC1. The net importers' 1990-2007 rises are distinguishable from zero at 10 percent (1990 sign) and 5 percent (1997 sign). |
| Excluded countries' and United States' shares of the top group's territorial CO2 (analysis sample) | Oil exporters plus transition economies: 6.0% (1990, 38 members), 5.8% (1997), 7.6% (2007), 10.4% (2017), 11.0% (2022); with Norway 6.0, 5.8, 7.9, 10.8, 11.4. United States: 43.9, 44.6, 43.7, 41.0, 42.7%. Denominator: the summed co2 of the 1990 top tercile's members in the analysis sample in each year. | The core exclusion removes a small and rising share; the United States dominates every weighted and aggregate figure. |
| Top-tercile wedge elasticity, 2007-2022 (all), and weighted estimates | 0.1866 (0.2128), wild p = 0.5005; weighted 1990-2007 0.0665 (0.1031), wild p = 0.5634; 2007-2022 -0.0432 (0.0332), wild p = 0.1495; Kish n 4.55 and 4.67; weighted minus unweighted 2007-2022 -0.2298 (bootstrap SE 0.2819), DuMouchel-Duncan p = 0.5531; without Malta -0.0075 (HC3 p 0.8961, wild p 0.8847), weighted -0.0465, weighted minus unweighted -0.0391 (bootstrap SE 0.0627), DuMouchel-Duncan p = 0.7762 | Inconclusive; the size of the unweighted slope, and its gap from the weighted one, are Malta's. |
| Growth share of the mean wedge change (point and interval) | 1990-2007: 14.5% [-107.3%, 136.3%]; 2007-2022: 40.3% [-49.8%, 130.5%] | Almost uninformative. |
| Benchmark-year drift, top tercile | 1997-2007 aggregate 0.0604, weighted intercept 0.0640 (0.0152); 2007-2017 aggregate -0.0209, mean 0.0233 (0.0433); 2007-2019 aggregate -0.0117; 2017-2022 aggregate 0.0530 | If the benchmark dating is right, the post-2007 rise is confined to extended-forward years. |
| File-internal test of the benchmark years | Kink statistic: benchmark mean 0.0510 vs 0.0599, permutation p = 0.8307; dispersion 0.1413 vs 0.1441, p = 0.5401; five largest kinks in 1999, 2000, 2008, 2009 and 1998 | The test's power is unknown and its largest kinks coincide with the 1998-2000 and 2008-2009 shocks, so its null neither confirms nor refutes the benchmark dating. |
| Kuwait variants and levels regressions | Long difference 1990-2007 with Kuwait 1990 kept: 0.0020 (0.0884); exploratory annual run 1.3181 (0.5787), without Kuwait 0.0799 (0.1291); levels with group-by-period FE 0.0815 (0.0790) and 0.0426 (0.0697) | The long-difference result does not depend on Kuwait; the levels design gives small, insignificant slopes. |
Caveats
- The data were supplied by the user, who describes them as Our World in Data's CO2 dataset, and were not independently verified (sha256 7ec10de1...).
- Conditional associations, not causal effects; there is no source of exogenous variation, and resource-led, services-led and terms-of-trade-driven growth are not separated.
- Power is low: 27-40 countries per group, minimum detectable effects 0.33-1.88; an insignificant slope is not evidence of no displacement.
- All composition variants (core, core without Norway, 1997 and 1990 net importers) were defined after the main results were known and are diagnostic. The oil-exporter set rests on an outside criterion that the file cannot check; no rule from the file's columns reproduces it. The 1997-sign net-importer group is selected inside 1990-2007; the 1990-sign group avoids this.
- One inference rule applies to the top group's long-difference wedge slopes in Tables 3-6: wild-bootstrap p below 0.10. Four of 20 distinct slopes in Table 4 pass it, from two pairs of overlapping groups, each with an HC3 p-value of 0.21-0.31; none passes in Tables 5 or 6. All four still pass after deleting their most influential country, but each one's size is highly sensitive to that country, and both deletions were chosen after the fact: without Malta the core group's 2007-2017 slopes are 0.304 and 0.339 (HC3 p 0.091, 0.066; wild p 0.068, 0.044); without the United Arab Emirates the net-importer slopes rise to 0.512 and 0.562 (HC3 p 0.059, 0.028; wild p 0.069, 0.036), so surviving that deletion says little about robustness: the net-importer passes survive only 15 of 27 and 5 of 28 single-country deletions, and the 1990-sign pass fails without Malta (wild p 0.1035). Wherever one country dominates the robust variance term, as Malta does in every unweighted sample starting in 2007 that includes it (the sample restricted to populations of at least 1 million starts in 2007 but excludes it), the wild-bootstrap p may be unreliable; its size in that case was not examined. About 2 passes are expected under the null and none survives Bonferroni or Holm correction. Intercepts, territorial and consumption slopes, and annual and levels specifications are outside the rule.
- P2 (a smaller slope after 2007) is informative only if P1 holds, and P1 is unsupported. In the four groups with both periods the 2007-2022 slope exceeds the 1990-2007 slope only because of Malta; without it, a deletion chosen after the fact, the 1990-2007 versus 2007-2022 ordering reverses in every group, while between the 1997-2007 and 2007-2017 benchmarks it is mixed (not reversed in the full group, about equal in the core group, reversed for the 1997 net importers); the small Malta-deleted 2007-2017 slopes of the full group and the 1997 net importers are themselves sensitive to the United Arab Emirates (Cook's-distance screen: 0.228 and 0.344, wild p 0.037 and 0.032); stacked tests distinguish the periods neither with Malta (p 0.44-0.52) nor without (0.39-0.80).
- Aggregate wedge changes are descriptive totals with no sampling error attached; about 41-45 percent of the group's territorial CO2 is the United States, and the weighted regressions have effective samples of about 4-10 countries.
- consumption_co2 is model-based. The benchmark years come from the Global Carbon Budget 2025 supplement and presume that the file is that release, which is not verified; the file-internal test has unknown power and its largest kinks coincide with the 1998-2000 and 2008-2009 shocks, so it neither confirms nor refutes the dating.
- Outside benchmark years the annual consumption series is built from GDP and trade series, so annual first-difference slopes are descriptive only.
- The Kuwait 1990-1992 exclusion was decided after an exploratory run that kept Kuwait; it rests on measurement grounds (1991 co2 492.98 Mt against 37.86 Mt in 1990; GDP fell 71.4 percent), and all variants are reported.
- The leave-one-out re-estimates, the Malta deletions, the wild-bootstrap p-values for Tables 5 and 6, the multiple-testing corrections and the Cook's-distance screen were added after the main results were known and are diagnostic; the screen is a sensitivity check only.
6.Paper writing
Manuscript drafted: “Decoupling or Displacement? The GDP Elasticity of Consumption-Based versus Territorial CO₂ Emissions, 1990–2022” (≈ 11320 words).
≈ 11320 words · AI-drafted manuscript · not peer-reviewed
Decoupling or Displacement? The GDP Elasticity of Consumption-Based versus Territorial CO₂ Emissions, 1990–2022
Abstract
If rich countries decoupled through carbon imports, consumption-based CO₂ should be more GDP-elastic than territorial CO₂. On 119 countries, the top income tercile's 1990–2007 consumption–territorial wedge elasticity is 0.027 (0.117); post hoc variants (without oil exporters and transition economies; 1990 net carbon importers) give 0.305 (0.234) and 0.105 (0.120). None is significant; sizeable displacement cannot be excluded. Dropping Malta post hoc leaves 2007–2022 slopes small and insignificant; no group's slope differs significantly between periods. Aggregate wedges rose 0.052–0.067 log points over 1990–2007 and fell 0.018–0.040 between the 2007 and 2017 benchmarks, significantly in weighted regressions only for 1997 net importers.
JEL: Q54, Q56, F18, O44
Keywords: decoupling; consumption-based emissions; carbon leakage; emissions embodied in trade; income elasticity; long differences
I. Introduction
Territorial CO₂ emissions have fallen in many high-income economies while GDP kept growing (Le Quéré et al. 2019). Absolute decoupling of this kind is rare and slow compared with climate targets (Haberl et al. 2020), and less common on a consumption basis: Hubacek et al. (2021) count 32 of 116 countries decoupling production-based emissions in 2015–2018 but only 23 decoupling consumption-based emissions, and Vogel and Hickel (2023) find 11 high-income countries achieving absolute consumption-based decoupling over 2013–2019, far below Paris-consistent rates. Karstensen, Peters, and Andrew (2018) trace the European Union's two accounts over 1990–2016. These studies count episodes or describe trends. They do not ask whether the gap between the two accounts scales with growth, which is the signature of displacement: rich economies growing by importing carbon-intensive goods, so that emissions transferred through trade rise (Davis and Caldeira 2010; Peters et al. 2011; Grubb et al. 2022).
The early income–emissions literature asked its question through the environmental Kuznets curve (Grossman and Krueger 1995), but levels regressions of log emissions on log income are fragile to functional form, nonstationarity and cross-sectional dependence (Stern 2004, 2017; Wagner 2008). We therefore use long differences. Write consumption-based emissions as territorial emissions plus net embodied imports. If displacement is how rich countries decouple, the GDP elasticity of the wedge should be positive, and it should have shrunk after net emission transfers peaked in the mid-2000s (Wood et al. 2020). We pre-specified both predictions, together with their refutation criteria, and estimate them by tercile of 1990 GDP per capita on a country-year panel supplied by the user as Our World in Data's CO₂ dataset.
The first prediction is not supported, but the data cannot reject it either. Over 1990–2007 the top-tercile wedge elasticity is 0.027 (0.117), with an HC3 standard error of 0.138 and a wild-bootstrap p-value of 0.81. An insignificant slope is not evidence of no displacement: the upper confidence bound, 0.258, times the group's mean growth of 0.640 gives 0.165 log points, more than the group's mean wedge increase of 0.121, and the minimum detectable effect at 5 percent size and 80 percent power is 0.33. Ranking by 1990 income admits 5 oil exporters and 6 transition economies (Appendix Table A1). In variants defined after the main results were known, the estimate is 0.305 (0.234) without them, in what we call the core group, and 0.105 (0.120) among the 28 net importers of embodied CO₂ in 1990; both intervals contain 0.027. We judge every long-difference wedge slope of the top group (Tables 3–6) by its wild-bootstrap p-value. In the four groups estimated for both periods, the 2007–2022 slopes (0.187–0.781) exceed the 1990–2007 slopes (0.027–0.305) only because of Malta. Without Malta, a deletion we chose after the fact, the 2007–2022 slopes are small and insignificant (−0.0075 to 0.158), the 1990–2007 slopes barely move and stay imprecise (0.026 to 0.297), and the ordering of these two periods reverses in every group. Between benchmark years, 1997–2007 against 2007–2017, the picture is mixed: the ordering reverses among the 1997 net importers, the core group's two slopes are about equal, and the full group's do not reverse. With or without Malta, stacked tests cannot distinguish the periods (all p > 0.3). The small Malta-deleted 2007–2017 slopes of the full group and the 1997 net importers are themselves sensitive to a second country, the United Arab Emirates: in these two samples, a sensitivity screen that drops each country with Cook's distance above 4/n, applied to every sample alike, removes exactly Malta and the Emirates and gives slopes of 0.228 (wild-bootstrap p = 0.037) and 0.344 (p = 0.032). Four of the 20 slopes in Table 4 pass at 10 percent: between the 2007 and 2017 benchmarks, the core group's (0.985, p = 0.068) and that of the core group without Norway (1.024, p = 0.051); between the 1997 and 2007 benchmarks, those of net importers selected by the 1997 sign (0.279, p = 0.091) and by the 1990 sign (0.294, p = 0.097). They come from two pairs of overlapping groups, and each has an HC3 p-value of 0.21–0.31. All four still pass when their most influential country is deleted, but the size of each is highly sensitive to that one country, and we chose both deletions after the fact. Without Malta the core slopes are 0.304 and 0.339 (HC3 p = 0.091 and 0.066; wild-bootstrap p = 0.068 and 0.044). Without the United Arab Emirates, an oil exporter removed from the core group, the net-importer slopes are 0.512 and 0.562 (HC3 p = 0.059 and 0.028; wild-bootstrap p = 0.069 and 0.036). Because this deletion raises both slopes and lowers their p-values, surviving it says little about robustness: the net-importer slopes pass in only 15 of 27 and 5 of 28 single-country deletions, and the 1990-sign slope fails without Malta (wild-bootstrap p = 0.104). Deletions chosen after the fact are not evidence for or against displacement, whichever way they move the slopes, and wherever one country dominates the robust variance term, as Malta does (64–87 percent) in every unweighted sample that starts in 2007 and includes it (the sample restricted to populations of at least 1 million starts in 2007 but excludes it), the wild-bootstrap p-value may be unreliable. About 2 of the 20 would pass if every null held, and none is significant after a Bonferroni or Holm correction.
Aggregate totals are descriptive but consistent across group definitions: the top tercile's aggregate wedge rose 0.064 log points over 1990–2007 (0.067 in the core group, 0.052 among 1990 net importers) and fell 0.021 between the 2007 and 2017 input–output benchmarks (0.019 in the core group). Consistency is not precision: in weighted regressions the full and core groups' 2007–2017 falls, −0.018 (0.012) and −0.020 (0.013), are not distinguishable from zero, and the United States emits 41–45 percent of the group's territorial CO₂. If the benchmark years are those listed in the Global Carbon Budget 2025 supplement, an outside assumption the file itself cannot confirm, the entire 2007–2022 rise falls in the extrapolated, pandemic-affected years 2017–2022.
Contribution. First, earlier comparisons of the two accounts used fixed-effects levels designs on small samples ending in 2008–2011. Knight and Schor (2014) studied 29 high-income countries over 1991–2008 and found a larger GDP effect on consumption-based emissions; Fernández-Amador et al. (2017) and Cohen et al. (2018) report related results. With the same design on 119 countries through 2022, adding group-by-period fixed effects, we find top-tercile wedge slopes of 0.082 (0.079) before 2008 and 0.043 (0.070) after, neither significant. Second, we separate a growth-linked slope from a common drift, weight by emissions to approximate the aggregate question, and show how group composition, endpoints and small-sample inference affect each conclusion. Trade-and-environment research decomposes emissions into scale, composition and technique effects (Copeland and Taylor 2004; Cherniwchan, Copeland, and Taylor 2017), and plant- and firm-level studies mostly find no marginal leakage (Naegele and Zaklan 2019; Dechezleprêtre et al. 2022) and a dominant technique effect (Levinson 2009; Shapiro and Walker 2018); our country-level tests lack the power to confirm or contradict the aggregate implications. The contribution is incremental and descriptive: unlike Aichele and Felbermayr (2012, 2015), who use Kyoto commitments as a policy shock, we have no exogenous variation.
Sections II–VI present the framework, data and strategy, results, robustness and conclusions. The Appendix lists group members, annual first differences, file-internal tests, emission shares, the rules tried for the oil-exporter set, leave-one-out estimates for the slopes that pass our inference rule, multiple-testing corrections and every slope without Malta.
II. Conceptual Framework
Let be territorial CO₂ and net CO₂ embodied in trade (imports minus exports), so . Define the wedge . Then
For income group and period , write . Growth-linked displacement means that demand shifts toward goods whose carbon-intensive stages are produced abroad as income rises, through the composition effect of Copeland and Taylor (2004). It implies . The intercept is a common drift: a change in the wedge shared by the group's members regardless of their own growth, as when global integration raises for all importers, which implies . Our predictions were P1: over 1990–2007, at the 5 percent level, and P2: smaller over 2007–2022. P2 is only informative if P1 holds. If is near zero before 2007, a point-estimate criterion for a decline is met roughly half the time under the null.
Identifying assumption. Equation (1) differences out shocks that move and proportionally. Within a tercile, is a causal growth slope only if growth differences are uncorrelated with other drivers of , such as resource discoveries, terms of trade or trade policy. We do not claim this; is a conditional association. Those other drivers are strongest in oil exporters and transition economies, so we also report the top group without them.
III. Data and Empirical Strategy
Data. The data were supplied by the user who commissioned this study as Our World in Data's CO₂ and greenhouse-gas emissions dataset (owid-co2-data.csv; Our World in Data n.d.). The file was received on October 29, 2026. It has 50,411 rows and 79 columns, and its SHA-256 checksum is 7ec10de1e9502c20c70bfff54476d7893922f7594aa299d9291d87fe49f3e342. We have not verified it against the public release. Definitions and sources follow the Our World in Data codebook, whose column names match the file. Under that codebook, territorial and consumption-based CO₂ come from the Global Carbon Budget 2025 (Friedlingstein et al. 2026), and GDP comes from the Maddison Project Database 2023 (Bolt and van Zanden 2025). We use four variables:
co2: territorial fossil and industrial CO₂, excluding land-use change, in Mt;consumption_co2: consumption-based CO₂, in Mt;gdp: real GDP in 2011 international dollars at purchasing-power parity;population: persons.
trade_co2 equals consumption_co2 minus co2 to within 0.0005 Mt.
How the consumption series is built. Per the Global Carbon Budget 2025 supplement (Friedlingstein et al. 2026, Sect. S1), detailed multi-regional input–output estimates exist only for the GTAP years 1997, 2001, 2004, 2007, 2011, 2014 and 2017; other years are extended from these benchmarks with GDP and trade time series (see also Peters, Davis, and Andrew 2012). On that dating, 1990–1995 are backcast, 2005 and 2006 lie between benchmarks, and 2019–2022 are extrapolated. The dating presumes that the file is the Global Carbon Budget 2025 release, which we have not verified, and the file does not corroborate it: annual wedge changes show neither sharper kinks nor greater cross-country dispersion at the stated benchmark years than in other years (permutation p = 0.83 and 0.54; Appendix Table A3). The test's power is unknown, and its largest kinks fall in 1998–2000 and 2008–2009, years of large global shocks, so its null neither confirms nor refutes the dating. The endpoint classification is therefore an outside assumption. Because extended 1990 and 2022 values use GDP and could mechanically link the wedge to growth, we also report benchmark-to-benchmark differences. For the same reason annual first-difference slopes largely reflect how the series is built; we report them only descriptively (Appendix Table A2).
Sample. We keep entities with an ISO code over 1990–2022 that have GDP, population and both emissions measures positive. Income terciles come from 1990 GDP per capita, with cut-offs of 4,695 and 12,349 international dollars; this gives 40 low, 39 middle and 40 high countries (Appendix Table A1). The top tercile includes 5 oil exporters (Kuwait, Venezuela, Trinidad and Tobago, the United Arab Emirates and Saudi Arabia) and 6 formerly centrally planned transition economies (Czechia, Slovakia, Slovenia, Estonia, Latvia and Lithuania); we chose these exclusions after the main results were known. The file cannot define the oil exporters: it has no production or export data, and every threshold on its one extraction-specific column, CO₂ from gas flaring, that keeps all five also selects Norway, Canada and Australia (Appendix Table A5). The set rests instead on an outside criterion, exports dominated by oil and gas. Norway meets it and ranks second of the 40 by mean flaring share, so leaving it in the core group was an inconsistency. Norway's consumption series begins only in 2003, so it is absent from differences that start before 2007; for later periods we report the core group with and without it. Net importers of embodied CO₂ (positive trade_co2) are 27 of the 39 members with 1997 values and 28 of the 38 with 1990 values.
Kuwait. Our baseline drops Kuwait for 1990–1992 (3 rows), leaving 119 countries and 3,905 country-years. We disclose that this exclusion was decided after an exploratory run that kept those years, in which the year-on-year wedge slope for the top tercile relative to other countries was 1.318 (0.579), clustered by country; without Kuwait the same regression gives 0.080 (0.129). The measurement case is that Kuwait's 1991 co2 in the file is 492.98 Mt against 37.86 Mt in 1990 and 29.68 Mt in 1992; the separate "Kuwaiti Oil Fires" entity has no co2 values, so the fires appear to be booked to Kuwait; and Kuwait's GDP fell 71.4 percent in 1991. Section V reports results with Kuwait kept and with only 1991–1992 dropped.
Estimation. The primary specification is a long difference over :
with tercile-specific intercepts and slopes and HC1 standard errors. For the top tercile we add HC3 standard errors and wild-bootstrap p-values (Webb weights, null imposed, 9,999 draws). For the top group's long-difference wedge slopes, unweighted and weighted, in Tables 3–6, one inference rule applies in both directions: a slope is significant at 10 percent only if its wild-bootstrap p-value is below 0.10. Intercepts, slopes of territorial and consumption-based emissions, and the annual and levels specifications are outside the rule; their stars use the stated standard errors. For the slopes that pass the rule we re-estimate leaving out one country at a time and report leverage, studentized residuals, Cook's distance and each country's share of the robust variance term (Appendix Table A6) and correct for multiple testing with the Bonferroni and Holm methods (Appendix Table A7). Where one country dominates the robust variance term, as Malta does in every unweighted sample starting in 2007 that includes it (Appendix Table A8; the sample restricted to populations of at least 1 million starts in 2007 but excludes it), the wild-bootstrap p-value may be unreliable, as its size in that case is unexamined. We report the intercepts in two forms: with demeaned within tercile, so the intercept is the drift at the tercile's mean growth, and in raw form, as the drift at zero growth. For the top tercile we also estimate (2) by weighted least squares with base-year co2 as weights. This approximates the aggregate question of whether the rich world's decoupling is displaced, while the unweighted estimand describes the average rich country. Its Kish effective sample is only about 4.6 countries, so weighted inference is fragile. We test the difference between weighted and unweighted slopes with a DuMouchel–Duncan test (HC3) and a country-level pairs bootstrap (4,999 draws). Stacking the two periods,
with country-clustered standard errors, tests . Annual first differences (Appendix) are
with and group-by-period intercepts , using Driscoll and Kraay (1998) standard errors with two lags alongside clustered ones. To mirror earlier studies, the levels equation is
with and group-by-period fixed effects ; the original specification omits , and a variant replaces it with country linear trends. Standard errors are clustered by country.
IV. Results
Net imports and drift. Table 1 shows that the top tercile's aggregate net CO₂ imports rose from 863.5 Mt (7.39 percent of territorial emissions) in 1990 to 2,019.4 Mt (14.43 percent) in 2007 and 2,149.0 Mt (18.16 percent) in 2022.
Table 1—Net CO₂ Embodied in Trade, Top Tercile of 1990 GDP per Capita
| 1990 | 2007 | 2022 | |
|---|---|---|---|
| Countries | 38 | 40 | 40 |
Aggregate trade_co2 (Mt) | 863.5 | 2,019.4 | 2,149.0 |
Aggregate trade_co2 / co2 (%) | 7.39 | 14.43 | 18.16 |
Mean / median trade_co2_share (%) | 19.23 / 14.05 | 36.77 / 24.30 | 70.54 / 24.37 |
| Share of net importers | 0.737 | 0.825 | 0.825 |
Notes: Top-tercile countries with an ISO code. Aggregates are sums over countries. Norway and Kuwait lack 1990 values.
Table 2 reports the drift. For the average top-tercile country the wedge rose 0.121 (0.040) log points over 1990–2007 and 0.104 (0.048) over 2007–2022; the medians are 0.061 and 0.076, so a few countries' wedges rose sharply. Weighting by emissions reduces the drift to 0.073 (0.016) and 0.037 (0.010), and the balanced aggregate wedge rose 0.064 and 0.032. On benchmark endpoints the pre-2007 drift survives: the aggregate rose 0.060 over 1997–2007, with a weighted intercept of 0.064 (0.015). The post-2007 drift does not. Over 2007–2017 the aggregate fell 0.021, the country mean was 0.023 (0.043) with median −0.005, and the weighted intercept was −0.018 (0.012); over 2007–2019 the aggregate fell 0.012. The rise is concentrated in 2017–2022 (aggregate 0.053). On the stated dating those values are extrapolated and span the pandemic, so the post-2007 drift at baseline endpoints is not evidence of continuing displacement. Absolute decoupling is less frequent on a consumption basis: 8 rather than 11 of 38 top-tercile countries over 1990–2007, and 27 rather than 32 of 40 over 2007–2022.
Table 2—Mean Changes and Tercile Intercepts (Drift), Long Differences
| Period | Group | Drift at mean growth | Median | Drift at zero growth | |||
|---|---|---|---|---|---|---|---|
| 1990–2007 | Low | 0.753 | 0.702 | 0.860 | 0.158** (0.065) | 0.059 | 0.355* (0.207) |
| Middle | 0.735 | 0.327 | 0.289 | −0.038 (0.098) | 0.007 | 0.208* (0.116) | |
| High | 0.640 | 0.166 | 0.287 | 0.121*** (0.040) | 0.061 | 0.103 (0.083) | |
| High, weighted | 0.506 | 0.073*** (0.016) | 0.039 (0.057) | ||||
| High, aggregate | 0.064 | ||||||
| 2007–2022 | Low | 0.717 | 0.826 | 0.805 | −0.020 (0.077) | −0.030 | −0.010 (0.241) |
| Middle | 0.411 | 0.187 | 0.360 | 0.172* (0.104) | 0.048 | 0.042 (0.188) | |
| High | 0.225 | −0.224 | −0.120 | 0.104** (0.048) | 0.076 | 0.062 (0.040) | |
| High, weighted | 0.214 | 0.037*** (0.010) | 0.046*** (0.012) | ||||
| High, aggregate | 0.032 | ||||||
| 1997–2007 (benchmarks) | High | 0.450 | 0.136 | 0.256 | 0.121*** (0.034) | 0.093 | 0.095 (0.063) |
| High, weighted | 0.064*** (0.015) | 0.021 (0.051) | |||||
| High, aggregate | 0.060 | ||||||
| 2007–2017 (benchmarks) | High | 0.159 | −0.110 | −0.087 | 0.023 (0.043) | −0.005 | −0.046 (0.038) |
| High, weighted | −0.018 (0.012) | −0.030* (0.017) | |||||
| High, aggregate | −0.021 | ||||||
| 2007–2019 | High | 0.192 | −0.150 | −0.113 | 0.037 (0.046) | −0.001 | −0.065* (0.037) |
| High, weighted | −0.011 (0.011) | −0.054 (0.037) | |||||
| High, aggregate | −0.012 | ||||||
| 2017–2022 | High | 0.066 | −0.114 | −0.033 | 0.081*** (0.020) | 0.088 | 0.087*** (0.023) |
| High, weighted | 0.052*** (0.009) | 0.066*** (0.008) | |||||
| High, aggregate | 0.053 |
Notes: . Drift is the intercept of equation (2) for , with demeaned within tercile (column 6) or raw (column 8). HC1 standard errors are in parentheses. "Weighted" rows use base-year co2 weights, with weighted mean . "Aggregate" is over top-tercile countries observed at both endpoints (38 for 1990–2007, 39 for 1997–2007, 40 otherwise). Benchmark (1997, 2007, 2017) and extrapolated (2019, 2022) years follow the Global Carbon Budget 2025 supplement, presuming the file is that release. Benchmark-year rows were added after the main results were known and are diagnostic. * p<0.10, ** p<0.05, *** p<0.01.
Growth slopes. Table 3 reports equation (2). Over 1990–2007 the top-tercile wedge elasticity is 0.027 (0.117), HC3 standard error 0.138, wild-bootstrap p = 0.809, so P1 is not supported. Yet the upper bound times mean growth (0.258 × 0.640 ≈ 0.165) exceeds the mean drift of 0.121. Growth accounts for 14.5 percent of the average country's wedge change at the point estimate but for [−107, 136] percent over the slope's interval; the minimum detectable effect is 0.329. The weighted estimate is 0.067 (0.103), HC3 standard error 0.125. Over 2007–2022 the unweighted estimate is 0.187 (0.213), HC3 standard error 0.280, but −0.0075 without Malta. The weighted estimate is −0.043 (0.033), HC3 standard error 0.038, wild-bootstrap p = 0.15. Their difference, −0.230, has a pairs-bootstrap standard error of 0.282 and a DuMouchel–Duncan p-value of 0.553: the data do not show that the average rich country and the rich-world aggregate diverge after 2007. The stacked change, 0.159 (0.248), or −0.033 (0.129) without Malta, cannot distinguish the periods. The middle tercile's 1990–2007 wedge slope, −0.334 (0.135), is the only significant one of the six tercile-period wedge slopes; its 2007–2022 counterpart, 0.319 (0.635), does not replicate it, and it is a description, not an identified mechanism.
Table 3—Long-Difference GDP Elasticities
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
| Period | 1990–2007 | 2007–2022 | ||||
| × Low | 0.879*** (0.295) | 0.616** (0.242) | −0.262 (0.249) | 1.131*** (0.367) | 1.117*** (0.229) | −0.014 (0.372) |
| × Middle | 0.748*** (0.126) | 0.415*** (0.103) | −0.334** (0.135) | 1.001*** (0.140) | 1.320** (0.645) | 0.319 (0.635) |
| × High | 0.746*** (0.179) | 0.773*** (0.112) | 0.027 (0.117) | 0.452** (0.228) | 0.639** (0.317) | 0.187 (0.213) |
| High, HC3 SE / wild-bootstrap p | 0.138 / 0.809 | 0.280 / 0.500 | ||||
| High, 95% CI (HC1) | [−0.203, 0.258] | [−0.231, 0.604] | ||||
| High, emissions-weighted | 0.067 (0.103) | −0.043 (0.033) | ||||
| High, weighted HC3 SE / wild-bootstrap p | 0.125 / 0.563 | 0.038 / 0.150 | ||||
| High, weighted Kish effective n | 4.55 | 4.67 | ||||
| Weighted minus unweighted (bootstrap SE) | 0.039 (0.124) | −0.230 (0.282) | ||||
| DuMouchel–Duncan test, p-value | 0.907 | 0.553 | ||||
| Countries | 116 | 116 | 116 | 119 | 119 | 119 |
| R² | 0.512 | 0.384 | 0.119 | 0.681 | 0.451 | 0.038 |
Notes: Equation (2) with tercile intercepts and slopes and no other controls. HC1 standard errors are in parentheses. HC3 standard errors and wild-bootstrap p-values (Webb weights, null imposed, 9,999 draws) use the top-tercile subsample. Weighted rows use WLS with base-year co2 weights. The weighted-minus-unweighted standard error is from a country-level pairs bootstrap (4,999 draws); the DuMouchel–Duncan test adds the weight and its interaction with to the unweighted regression (HC3). The stacked high-tercile change (equation 3, 235 country-periods, clustered by country) is 0.159 (0.248); Appendix Table A8 gives the slopes without Malta. Top-tercile wedge slopes are judged by their wild-bootstrap p-values. HC3, bootstrap and difference tests were added after the main results were known. * p<0.10, ** p<0.05, *** p<0.01.
Group composition. Table 4 re-estimates the wedge slope and drift for the full group and four variants, all defined after the main results were known; stars on slopes follow the wild-bootstrap rule. The core group's 1990–2007 slope, 0.305 (0.234), has a 95 percent interval of [−0.152, 0.763] that contains the full-group 0.027, an HC3 standard error of 0.295 and a wild-bootstrap p-value of 0.337; its upper bound times mean growth is 0.469, far above the mean drift of 0.121, and its minimum detectable effect is 0.655, so it is too imprecise to distinguish from the full-group estimate. Dropping Norway as well changes nothing before 2007 and little after: 0.781 (0.542) rather than 0.738 (0.551) over 2007–2022.
The 1997 net importers are selected on the 1997 sign of net embodied trade, and hence of the wedge; 1997 lies inside 1990–2007, so selection affects that period's estimates too. Selected on the 1990 sign, net importers give 0.105 (0.120) over 1990–2007 (wild-bootstrap p = 0.322), close to the 1997-sign 0.130 (0.118). Between the 1997 and 2007 benchmarks the net-importer slopes, 0.279 (0.150) and 0.294 (0.160) by the 1997 and 1990 signs, have wild-bootstrap p-values of 0.091 and 0.097, below 0.10, though HC3 p-values of 0.28 and 0.31. The rule also works the other way: the core group's 2007–2017 slope, 0.985 (0.663), has an HC1 p-value of 0.137 but a wild-bootstrap p-value of 0.068 (1.024 and 0.051 without Norway; HC3 p = 0.22 and 0.21). No other slope, weighted or not, passes the rule. The four passing slopes come from two pairs of overlapping groups: the core group with and without Norway, and net importers by the two signs, which differ by three countries.
Single-country influence. In each of the 11 distinct unweighted samples that start in 2007 and include Malta (2005–2007 for averaged endpoints), Malta dominates (Appendix Table A8): it has the largest studentized residual (3.90–5.25), a Cook's distance of 0.79–2.25 against 4/n of 0.10–0.15, and 64–87 percent of the robust variance term . The sample restricted to populations of at least 1 million also starts in 2007 but excludes Malta. We chose to delete Malta after the fact, on the same terms as the United Arab Emirates below. Without Malta the 2007–2022 slopes of the full group, the core group, the core without Norway and the 1997 net importers are −0.0075, 0.102, 0.158 and 0.021, against 1990–2007 slopes of 0.026, 0.297, 0.297 and 0.128, all with HC3 and wild-bootstrap p-values of 0.20 or more. Between these two periods the ordering reverses in every group, yet no stacked test (equation 3) distinguishes them: p = 0.44–0.52 with Malta and 0.39–0.80 without. On the benchmark endpoints the picture without Malta is mixed. The 1997–2007 and 2007–2017 slopes are 0.074 and 0.125 in the full group, not reversed; 0.348 and 0.304 in the core group and 0.348 and 0.339 without Norway, about equal; and 0.298 and 0.144 among the 1997 net importers, reversed (stacked p ≥ 0.35 throughout). The small, insignificant 2007–2017 slopes of the full group and the 1997 net importers are themselves sensitive to the United Arab Emirates. In these two samples, the Cook's-distance screen described under Appendix Table A8 drops exactly Malta and the Emirates, and the slopes become 0.228 (HC3 p = 0.033; wild-bootstrap p = 0.037) and 0.344 (HC3 p = 0.049; wild-bootstrap p = 0.032). Nor are all post-2007 slopes small without Malta: the core group's 2007–2017 slopes, 0.304 and 0.339, still pass our rule (below). The screen is a sensitivity check and is itself a deletion rule; deletions chosen after the fact, in either direction, are not evidence. Malta barely affects the weighted regressions.
Each of the four slopes passes in the full sample and still passes when its most influential country is deleted, but its size is highly sensitive to that one country (Appendix Table A6); we report both deletions on the same terms, as both were chosen after the fact. Without Malta the core group's 2007–2017 slopes fall to 0.304 and 0.339, about its 1990–2007 estimate, with HC3 p = 0.091 and 0.066 and wild-bootstrap p = 0.068 and 0.044. The United Arab Emirates, an oil exporter removed from the core group but a net importer by both signs, has a leverage of 0.585 and 0.574 (mean about 0.07) and 44 and 54 percent of the robust variance term. Without it the net-importer slopes rise to 0.512 and 0.562, with HC3 p = 0.059 and 0.028 and wild-bootstrap p = 0.069 and 0.036. Neither set of slopes is evidence for or against displacement, and with one country dominating, the full-sample wild-bootstrap p-values of all four may be unreliable. Other deletions move the slopes much less but leave the net importers' wild-bootstrap p-values below 0.10 in only 15 of 27 and 5 of 28 fits.
Multiple testing. Table 4 has 20 distinct unweighted slopes. If every null hypothesis held, about 2 would pass at 10 percent; 4 do. If the tests were independent, the probability of 4 or more passes would be 0.133, but the periods are nested and the samples overlap, so this is only a reference. The Bonferroni threshold at 10 percent is 0.10/20 = 0.005, far below the smallest wild-bootstrap p-value of 0.051, and Holm's step-down correction rejects nothing either (Appendix Table A7): no slope is significant familywise. Adding the long-difference slopes of Tables 5 and 6 gives 29 tests, still with 4 passes against 2.9 expected; with Malta deleted from every Table 4 sample, 3 of 20 pass (smallest p-value 0.044).
The aggregate wedge changes are descriptive totals, consistent across group definitions: over 1990–2007 they lie between 0.052 and 0.067 in all five groups, and over 2007–2017 between −0.018 and −0.040 in the four estimated; removing Malta changes none by more than 0.002. Beside the 2007–2017 falls, the weighted intercepts of the full group, the core group and the core without Norway are −0.018 (0.012), −0.020 (0.013) and −0.020 (0.012), on Kish effective samples of 4.7, 4.0 and 4.0 countries; none is distinguishable from zero, before allowing for inventory error. Only the 1997 net importers' fall, −0.043 (0.016) on an effective sample of 10.1, differs from zero under HC1. The 11 excluded countries emit 6.0 percent of the group's territorial CO₂ in 1990, 5.8 in 1997, 7.6 in 2007, 10.4 in 2017 and 11.0 in 2022, and the United States 41.0–44.6 percent (Appendix Table A4). Both net-importer groups exclude the United States and Australia and show the pre-2007 rise, and their precision deserves the same report: their 1990–2007 weighted intercepts are 0.050 (0.028, p = 0.070) by the 1990 sign and 0.059 (0.028, p = 0.035) by the 1997 sign, the first distinguishable from zero only at 10 percent.
Table 4—Top-Group Composition: Wedge Elasticity and Drift
| Sample | Period | N | Wedge elasticity (HC1) | HC3 SE | Wild p | 95% CI (HC1) | Weighted (HC1) [HC3 SE] | Weighted wild p | Kish n | Mean | Weighted drift | Aggregate |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| All top tercile | 1990–2007 | 38 | 0.027 (0.117) | 0.138 | 0.809 | [−0.203, 0.258] | 0.067 (0.103) [0.125] | 0.563 | 4.55 | 0.121*** (0.040) | 0.073*** (0.016) | 0.064 |
| 2007–2022 | 40 | 0.187 (0.213) | 0.280 | 0.500 | [−0.231, 0.604] | −0.043 (0.033) [0.038] | 0.150 | 4.67 | 0.104** (0.048) | 0.037*** (0.010) | 0.032 | |
| 1997–2007 | 39 | 0.057 (0.120) | 0.126 | 0.613 | [−0.178, 0.293] | 0.137 (0.139) [0.182] | 0.504 | 4.47 | 0.121*** (0.034) | 0.064*** (0.015) | 0.060 | |
| 2007–2017 | 40 | 0.437 (0.344) | 0.379 | 0.213 | [−0.238, 1.112] | 0.080 (0.090) [0.110] | 0.456 | 4.67 | 0.023 (0.043) | −0.018 (0.012) | −0.021 | |
| 2017–2022 | 40 | −0.096 (0.089) | 0.281 | 0.558 | [−0.270, 0.078] | −0.195 (0.040) [0.088] | 0.189 | 5.17 | 0.081*** (0.020) | 0.052*** (0.009) | 0.053 | |
| Core: without oil exporters and transition economies | 1990–2007 | 28 | 0.305 (0.234) | 0.295 | 0.337 | [−0.152, 0.763] | 0.203 (0.168) [0.210] | 0.361 | 4.03 | 0.121*** (0.043) | 0.073*** (0.016) | 0.067 |
| 2007–2022 | 29 | 0.738 (0.551) | 0.677 | 0.213 | [−0.341, 1.817] | −0.052 (0.083) [0.095] | 0.530 | 4.01 | 0.121** (0.061) | 0.036*** (0.010) | 0.038 | |
| 1997–2007 | 28 | 0.325 (0.342) | 0.378 | 0.365 | [−0.346, 0.996] | 0.377 (0.225) [0.403] | 0.468 | 3.98 | 0.133*** (0.041) | 0.065*** (0.013) | 0.064 | |
| 2007–2017 | 29 | 0.985* (0.663) | 0.800 | 0.068 | [−0.315, 2.285] | 0.052 (0.163) [0.187] | 0.780 | 4.01 | 0.034 (0.054) | −0.020 (0.013) | −0.019 | |
| 2017–2022 | 29 | −0.561 (0.421) | 0.492 | 0.255 | [−1.386, 0.265] | −0.342 (0.170) [0.267] | 0.140 | 4.21 | 0.087*** (0.020) | 0.056*** (0.010) | 0.057 | |
| Core, also without Norway | 2007–2022 | 28 | 0.781 (0.542) | 0.670 | 0.121 | [−0.282, 1.844] | −0.038 (0.082) [0.094] | 0.645 | 3.98 | 0.142** (0.059) | 0.038*** (0.010) | 0.040 |
| 2007–2017 | 28 | 1.024* (0.671) | 0.816 | 0.051 | [−0.290, 2.339] | 0.062 (0.165) [0.190] | 0.746 | 3.98 | 0.042 (0.055) | −0.020 (0.012) | −0.018 | |
| 2017–2022 | 28 | −0.609 (0.405) | 0.474 | 0.157 | [−1.403, 0.185] | −0.342 (0.174) [0.284] | 0.168 | 4.18 | 0.100*** (0.016) | 0.058*** (0.010) | 0.058 | |
| Net importers of embodied CO₂, 1997 sign | 1990–2007 | 27 | 0.130 (0.118) | 0.138 | 0.183 | [−0.102, 0.363] | 0.172 (0.105) [0.124] | 0.176 | 8.70 | 0.142*** (0.047) | 0.059** (0.028) | 0.054 |
| 2007–2022 | 27 | 0.527 (0.499) | 0.598 | 0.485 | [−0.452, 1.506] | −0.089 (0.097) [0.120] | 0.455 | 10.14 | 0.137** (0.065) | 0.039* (0.020) | 0.044 | |
| 1997–2007 | 27 | 0.279* (0.150) | 0.260 | 0.091 | [−0.015, 0.572] | 0.308 (0.141) [0.219] | 0.222 | 9.48 | 0.130*** (0.037) | 0.050** (0.023) | 0.052 | |
| 2007–2017 | 27 | 0.626 (0.539) | 0.656 | 0.346 | [−0.431, 1.682] | −0.065 (0.113) [0.137] | 0.577 | 10.14 | 0.030 (0.061) | −0.043*** (0.016) | −0.040 | |
| 2017–2022 | 27 | −0.426 (0.418) | 0.490 | 0.531 | [−1.246, 0.394] | −0.047 (0.162) [0.201] | 0.775 | 9.90 | 0.106*** (0.017) | 0.080*** (0.012) | 0.084 | |
| Net importers of embodied CO₂, 1990 sign | 1990–2007 | 28 | 0.105 (0.120) | 0.130 | 0.322 | [−0.130, 0.340] | 0.122 (0.098) [0.119] | 0.293 | 8.54 | 0.121*** (0.044) | 0.050* (0.028) | 0.052 |
| 1997–2007 | 28 | 0.294* (0.160) | 0.291 | 0.097 | [−0.019, 0.606] | 0.377 (0.146) [0.248] | 0.166 | 9.12 | 0.147*** (0.033) | 0.060*** (0.020) | 0.064 |
Notes: Top-group slopes from equation (2), HC1 standard errors in parentheses. Stars on the two slope columns follow our rule, the wild-bootstrap p-value (Webb weights, null imposed, 9,999 draws, top-group subsample). Under HC1 the two 2007–2017 core slopes would be unstarred (p = 0.137 and 0.127), and six weighted slopes would be starred: all, 2017–2022; core with and without Norway, 2017–2022; core, 1997–2007; both net-importer groups, 1997–2007. Drift stars use HC1. Weights are base-year co2. Mean is the intercept with demeaned within the group; weighted drift is the weighted intercept at weighted mean growth. Aggregate is over members observed at both endpoints, a descriptive total. Core excludes the oil exporters and transition economies of Appendix Table A1; without Norway, its 1990–2007 and 1997–2007 rows equal the core rows. Appendix Table A6 gives leave-one-out results for the four starred slopes, Table A7 the multiple-testing corrections and Table A8 every slope without Malta, which dominates every unweighted 2007–2022 and 2007–2017 slope, all of them in samples that include it. All variants were defined after the main results were known; diagnostic. * p<0.10, ** p<0.05, *** p<0.01.
V. Robustness and Heterogeneity
Kuwait, influential countries and specification. Table 5 reports variants of the full top-tercile wedge elasticity. Keeping Kuwait's 1990 anchor gives a 1990–2007 long difference of 0.002 (0.088); Kuwait matters marginally for the levels specification, 0.171 (0.098). Under our rule no long-difference variant is significant: the wild-bootstrap p-values, in brackets, are all at least 0.32. Leave-one-country-out ranges are narrow except for the unweighted post-2007 slope, from −0.0075 without Malta to 0.375 without Venezuela, whose output collapse makes it a high-leverage point. Group-by-period fixed effects or country linear trends in the levels equation leave the slopes close to the original specification and insignificant.
Table 5—Top-Tercile Wedge Elasticity: Sample and Specification Variants
| Specification | 1990–2007 / pre-2008 | 2007–2022 / post-2007 |
|---|---|---|
| Baseline long difference | 0.027 (0.117) [0.809] | 0.187 (0.213) [0.500] |
| Leave-one-country-out range | [−0.053, 0.089] | [−0.0075, 0.375] |
| Emissions-weighted, leave-one-out range | [−0.006, 0.138] | [−0.062, −0.022] |
| Kuwait kept, all years | 0.002 (0.088) [0.980] | 0.187 (0.213) [0.500] |
| Kuwait, only 1991–1992 dropped | 0.002 (0.088) [0.980] | 0.187 (0.213) [0.500] |
| First period starts 1995 | 0.101 (0.096) [0.324] | — |
| Population ≥ 1 million | 0.034 (0.118) [0.767] | −0.009 (0.054) [0.859] |
| Regressor | 0.146 (0.210) [0.495] | 0.208 (0.209) [0.421] |
| Levels, country and year FE | 0.087 (0.083) | 0.057 (0.078) |
| Levels, plus group-by-period FE | 0.082 (0.079) | 0.043 (0.070) |
| Levels, plus country linear trends | 0.033 (0.106) | 0.045 (0.106) |
| Levels, FE, Kuwait kept | 0.171* (0.098) | 0.140 (0.094) |
Notes: The dependent variable is (or in levels). Long differences use HC1 standard errors, with wild-bootstrap p-values in brackets (Webb weights, null imposed, 9,999 draws, top-group subsample); by our rule none is significant. Levels regressions cluster by country and are outside the rule; their stars use the clustered standard errors. Leave-one-out ranges are point estimates over 119 fits. Variants other than the baseline were added after the main results were known and are diagnostic. * p<0.10, ** p<0.05, *** p<0.01.
Endpoints. Table 6 replaces single-year endpoints. Before 2007 the slope stays near zero: 0.046 (0.249) with three-year averages and 0.057 (0.120) between benchmarks. After 2007, estimates range from 0.096 (0.209) with averaged endpoints to 0.536 (0.275) with 2019 as the end year, with 0.437 (0.344) between the 2007 and 2017 benchmarks; Malta carries 72–73 percent of each one's variance term, and without it they are −0.091, 0.286 and 0.125 (Appendix Table A8). The 2019-endpoint slope's p-value of 0.051 is an HC1 p-value; its HC3 p-value is 0.081 and its wild-bootstrap p-value 0.272, so it is not significant under our rule. Weighted estimates in the post-2007 variants range from −0.076 to 0.238, with wild-bootstrap p-values of at least 0.35. Any statement about the aggregate is therefore endpoint-specific: the weighted estimate is −0.043 (0.033) with baseline endpoints but 0.238 (0.175) with a 2019 end year. Post-2007 growth-linked displacement is neither established nor excluded.
Table 6—Top-Tercile Long-Difference Elasticities under Alternative Endpoints
Endpoints (status of consumption_co2) | : wild p | , weighted | Weighted: wild p | N (high) | |||
|---|---|---|---|---|---|---|---|
| 1990–92 → 2005–07 (backcast → between) | 0.866*** (0.293) | 0.912*** (0.201) | 0.046 (0.249) | 0.882 | 0.093 (0.158) | 0.618 | 114 (37) |
| 1997 → 2007 (benchmark → benchmark) | 0.497*** (0.131) | 0.555*** (0.116) | 0.057 (0.120) | 0.613 | 0.137 (0.139) | 0.504 | 118 (39) |
| 2005–07 → 2020–22 (between → extrapolated) | 0.551*** (0.213) | 0.647** (0.298) | 0.096 (0.209) | 0.876 | −0.076 (0.057) | 0.354 | 118 (40) |
| 2007 → 2017 (benchmark → benchmark) | 0.613*** (0.212) | 1.050*** (0.229) | 0.437 (0.344) | 0.213 | 0.080 (0.090) | 0.456 | 119 (40) |
| 2007 → 2019 (benchmark → extrapolated) | 0.555** (0.217) | 1.091*** (0.195) | 0.536 (0.275) | 0.272 | 0.238 (0.175) | 0.601 | 119 (40) |
Notes: Top-tercile slopes from equation (2), HC1 standard errors. Three-year windows require all years. Status follows the Global Carbon Budget 2025 supplement (Friedlingstein et al. 2026), presuming the file is that release. The two wedge columns follow our inference rule: their wild-bootstrap p-values (Webb weights, null imposed, 9,999 draws, top-group subsample) are all at least 0.10, so they carry no stars; under HC1 the 2019-endpoint slope of 0.536 would be starred (p = 0.051; HC3 p = 0.081). The two benchmark-to-benchmark rows equal the full-group rows of Table 4. The and columns are outside the rule; their stars use HC1. Rows added after the main results were known; diagnostic. * p<0.10, ** p<0.05, *** p<0.01.
Remaining threats. We do not separate resource-led from services-led growth, which bias in opposite directions; the composition variants, defined after the main results were known, address this only in part. Technology-adjusted accounting would likely shrink the measured drift (Kander et al. 2015; Jiborn et al. 2018). We ran no unit-root or cross-sectional-dependence diagnostics (Pesaran 2007), no placebo outcomes, and no controls for openness or energy prices. With 27–40 countries per group, power is low.
VI. Conclusion
The pre-specified hypothesis is not supported, but the evidence cannot reject it. For the full top tercile the 1990–2007 wedge elasticity is 0.027, near zero and insignificant; yet the upper bound times mean growth exceeds the mean drift and the minimum detectable effect is 0.33. The core estimate, from a group defined after the main results were known, is 0.305 (interval [−0.152, 0.763], minimum detectable effect 0.655), too imprecise to distinguish from the full-group 0.027, so whether displacement scales with growth cannot be settled with these data. The larger 2007–2022 slopes are Malta's: without it they are small and insignificant, and the 1990–2007 slopes remain imprecise. We chose that deletion after the fact, like the deletion of the United Arab Emirates, and the Malta-deleted 2007–2017 slopes of the full group and the 1997 net importers are themselves sensitive to the Emirates: a Cook's-distance screen that drops both countries gives 0.228 and 0.344, each passing our rule, while the core group's 2007–2017 slopes pass without Malta alone. The ordering of 1990–2007 and 2007–2022 slopes reverses in every group without Malta, but between benchmark years, 1997–2007 against 2007–2017, the picture is mixed, and with or without Malta no stacked test distinguishes the periods; by the criterion of Section II, P2 is uninformative because P1 is unsupported. Under our wild-bootstrap rule 4 of the 20 top-group slopes in Table 4 pass at 10 percent: the core group's 2007–2017 slopes (0.985, p = 0.068; 1.024, p = 0.051 without Norway) and the net importers' 1997–2007 slopes (0.279, p = 0.091; 0.294, p = 0.097). They come from two pairs of overlapping groups, each has an HC3 p-value of 0.21–0.31, about 2 would pass if every null held, and none survives a multiple-testing correction. All four still pass when their most influential country is deleted, but the size of each is highly sensitive to that one country: without Malta the core slopes fall to 0.304 and 0.339 (HC3 p = 0.091 and 0.066; wild-bootstrap p = 0.068 and 0.044); without the United Arab Emirates the net-importer slopes rise to 0.512 and 0.562 (HC3 p = 0.059 and 0.028; wild-bootstrap p = 0.069 and 0.036). Because this deletion strengthens them, surviving it says little about robustness: the net-importer slopes pass in only 15 of 27 and 5 of 28 single-country deletions, and the 1990-sign slope fails without Malta (wild-bootstrap p = 0.104). We chose both deletions after the fact, and deletions in either direction are not evidence for or against displacement; wherever one country dominates the variance term, as Malta does in every unweighted sample that starts in 2007 and includes it (the sample restricted to populations of at least 1 million starts in 2007 but excludes it), the wild-bootstrap p-value may be unreliable. The weighted and unweighted estimates do not differ significantly, and the weighted regression rests on an effective sample of about 4.6 countries.
The aggregate wedge changes are descriptive totals, consistent across group definitions and within 0.002 of their values without Malta. The top tercile's aggregate wedge rose 0.064 log points over 1990–2007 and 0.060 between the 1997 and 2007 benchmarks; it fell 0.021 between the 2007 and 2017 benchmarks, and its 0.032 rise over 2007–2022 comes entirely from 2017–2022. The same pattern appears without oil exporters and transition economies and among net importers, which exclude the United States, but consistency is not precision: the full and core groups' 2007–2017 falls are not distinguishable from zero in weighted regressions, before allowing for inventory error. Territorial accounting thus overstated rich-world decoupling by a modest, growing margin before 2007, and we find no sign that the margin grew between the 2007 and 2017 benchmarks. Whether it widened after 2017 cannot be judged until new benchmark estimates replace the extrapolated values, and even the reading that the post-2007 rise falls in extrapolated years is conditional on a benchmark dating the file cannot confirm. These conclusions rest on a single user-supplied file that we have not verified, on model-based consumption inventories extended with GDP data, and on associations rather than causal effects. Policy shocks of the kind Aichele and Felbermayr (2015) exploit offer a route to identification.
References
Aichele, Rahel, and Gabriel Felbermayr. 2012. "Kyoto and the Carbon Footprint of Nations." Journal of Environmental Economics and Management 63 (3): 336–54. https://doi.org/10.1016/j.jeem.2011.10.005.
Aichele, Rahel, and Gabriel Felbermayr. 2015. "Kyoto and Carbon Leakage: An Empirical Analysis of the Carbon Content of Bilateral Trade." Review of Economics and Statistics 97 (1): 104–15. https://doi.org/10.1162/REST_a_00438.
Bolt, Jutta, and Jan Luiten van Zanden. 2025. "Maddison-Style Estimates of the Evolution of the World Economy: A New 2023 Update." Journal of Economic Surveys 39 (2): 631–71. https://doi.org/10.1111/joes.12618.
Cherniwchan, Jevan, Brian R. Copeland, and M. Scott Taylor. 2017. "Trade and the Environment: New Methods, Measurements, and Results." Annual Review of Economics 9: 59–85. https://doi.org/10.1146/annurev-economics-063016-103756.
Cohen, Gail, João Tovar Jalles, Prakash Loungani, and Ricardo Marto. 2018. "The Long-Run Decoupling of Emissions and Output: Evidence from the Largest Emitters." Energy Policy 118: 58–68. https://doi.org/10.1016/j.enpol.2018.03.028.
Copeland, Brian R., and M. Scott Taylor. 2004. "Trade, Growth, and the Environment." Journal of Economic Literature 42 (1): 7–71. https://doi.org/10.1257/.42.1.7.
Davis, Steven J., and Ken Caldeira. 2010. "Consumption-Based Accounting of CO₂ Emissions." Proceedings of the National Academy of Sciences 107 (12): 5687–92. https://doi.org/10.1073/pnas.0906974107.
Dechezleprêtre, Antoine, Caterina Gennaioli, Ralf Martin, Mirabelle Muûls, and Thomas Stoerk. 2022. "Searching for Carbon Leaks in Multinational Companies." Journal of Environmental Economics and Management 112: 102601. https://doi.org/10.1016/j.jeem.2021.102601.
Driscoll, John C., and Aart C. Kraay. 1998. "Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data." Review of Economics and Statistics 80 (4): 549–60. https://doi.org/10.1162/003465398557825.
Fernández-Amador, Octavio, Joseph F. Francois, Doris A. Oberdabernig, and Patrick Tomberger. 2017. "Carbon Dioxide Emissions and Economic Growth: An Assessment Based on Production and Consumption Emission Inventories." Ecological Economics 135: 269–79. https://doi.org/10.1016/j.ecolecon.2017.01.004.
Friedlingstein, Pierre, Michael O'Sullivan, Matthew W. Jones, Robbie M. Andrew, Dorothee C. E. Bakker, Judith Hauck, Peter Landschützer, et al. 2026. "Global Carbon Budget 2025." Earth System Science Data 18 (5): 3211–88. https://doi.org/10.5194/essd-18-3211-2026.
Grossman, Gene M., and Alan B. Krueger. 1995. "Economic Growth and the Environment." Quarterly Journal of Economics 110 (2): 353–77. https://doi.org/10.2307/2118443.
Grubb, Michael, Nino David Jordan, Edgar Hertwich, Karsten Neuhoff, Kasturi Das, Kaushik Ranjan Bandyopadhyay, Harro van Asselt, et al. 2022. "Carbon Leakage, Consumption, and Trade." Annual Review of Environment and Resources 47: 753–95. https://doi.org/10.1146/annurev-environ-120820-053625.
Haberl, Helmut, Dominik Wiedenhofer, Doris Virág, Gerald Kalt, Barbara Plank, Paul Brockway, Tomer Fishman, et al. 2020. "A Systematic Review of the Evidence on Decoupling of GDP, Resource Use and GHG Emissions, Part II: Synthesizing the Insights." Environmental Research Letters 15 (6): 065003. https://doi.org/10.1088/1748-9326/ab842a.
Hubacek, Klaus, Xiangjie Chen, Kuishuang Feng, Thomas Wiedmann, and Yuli Shan. 2021. "Evidence of Decoupling Consumption-Based CO₂ Emissions from Economic Growth." Advances in Applied Energy 4: 100074. https://doi.org/10.1016/j.adapen.2021.100074.
Jiborn, Magnus, Astrid Kander, Viktoras Kulionis, Hana Nielsen, and Daniel D. Moran. 2018. "Decoupling or Delusion? Measuring Emissions Displacement in Foreign Trade." Global Environmental Change 49: 27–34. https://doi.org/10.1016/j.gloenvcha.2017.12.006.
Kander, Astrid, Magnus Jiborn, Daniel D. Moran, and Thomas O. Wiedmann. 2015. "National Greenhouse-Gas Accounting for Effective Climate Policy on International Trade." Nature Climate Change 5 (5): 431–35. https://doi.org/10.1038/nclimate2555.
Karstensen, Jonas, Glen P. Peters, and Robbie M. Andrew. 2018. "Trends of the EU's Territorial and Consumption-Based Emissions from 1990 to 2016." Climatic Change 151 (2): 131–42. https://doi.org/10.1007/s10584-018-2296-x.
Knight, Kyle W., and Juliet B. Schor. 2014. "Economic Growth and Climate Change: A Cross-National Analysis of Territorial and Consumption-Based Carbon Emissions in High-Income Countries." Sustainability 6 (6): 3722–31. https://doi.org/10.3390/su6063722.
Le Quéré, Corinne, Jan Ivar Korsbakken, Charlie Wilson, Jale Tosun, Robbie Andrew, Robert J. Andres, Josep G. Canadell, et al. 2019. "Drivers of Declining CO₂ Emissions in 18 Developed Economies." Nature Climate Change 9 (3): 213–17. https://doi.org/10.1038/s41558-019-0419-7.
Levinson, Arik. 2009. "Technology, International Trade, and Pollution from US Manufacturing." American Economic Review 99 (5): 2177–92. https://doi.org/10.1257/aer.99.5.2177.
Naegele, Helene, and Aleksandar Zaklan. 2019. "Does the EU ETS Cause Carbon Leakage in European Manufacturing?" Journal of Environmental Economics and Management 93: 125–47. https://doi.org/10.1016/j.jeem.2018.11.004.
Our World in Data. n.d. "CO₂ and Greenhouse Gas Emissions" dataset (owid-co2-data.csv) and codebook (owid-co2-codebook.csv). Codebook sources: Global Carbon Budget (2025); Bolt and van Zanden, Maddison Project Database 2023; Population based on various sources (2024). https://github.com/owid/co2-data (codebook accessed October 29, 2026).
Pesaran, M. Hashem. 2007. "A Simple Panel Unit Root Test in the Presence of Cross-Section Dependence." Journal of Applied Econometrics 22 (2): 265–312. https://doi.org/10.1002/jae.951.
Peters, Glen P., Steven J. Davis, and Robbie Andrew. 2012. "A Synthesis of Carbon in International Trade." Biogeosciences 9 (8): 3247–76. https://doi.org/10.5194/bg-9-3247-2012.
Peters, Glen P., Jan C. Minx, Christopher L. Weber, and Ottmar Edenhofer. 2011. "Growth in Emission Transfers via International Trade from 1990 to 2008." Proceedings of the National Academy of Sciences 108 (21): 8903–08. https://doi.org/10.1073/pnas.1006388108.
Shapiro, Joseph S., and Reed Walker. 2018. "Why Is Pollution from US Manufacturing Declining? The Roles of Environmental Regulation, Productivity, and Trade." American Economic Review 108 (12): 3814–54. https://doi.org/10.1257/aer.20151272.
Stern, David I. 2004. "The Rise and Fall of the Environmental Kuznets Curve." World Development 32 (8): 1419–39. https://doi.org/10.1016/j.worlddev.2004.03.004.
Stern, David I. 2017. "The Environmental Kuznets Curve after 25 Years." Journal of Bioeconomics 19 (1): 7–28. https://doi.org/10.1007/s10818-017-9243-1.
Vogel, Jefim, and Jason Hickel. 2023. "Is Green Growth Happening? An Empirical Analysis of Achieved versus Paris-Compliant CO₂–GDP Decoupling in High-Income Countries." Lancet Planetary Health 7 (9): e759–69. https://doi.org/10.1016/S2542-5196(23)00174-2.
Wagner, Martin. 2008. "The Carbon Kuznets Curve: A Cloudy Picture Emitted by Bad Econometrics?" Resource and Energy Economics 30 (3): 388–408. https://doi.org/10.1016/j.reseneeco.2007.11.001.
Wood, Richard, Michael Grubb, Annela Anger-Kraavi, Hector Pollitt, Ben Rizzo, Eva Alexandri, Konstantin Stadler, et al. 2020. "Beyond Peak Emission Transfers: Historical Impacts of Globalization and Future Impacts of Climate Policies on International Emission Transfers." Climate Policy 20 (sup1): S14–S27. https://doi.org/10.1080/14693062.2019.1619507.
Appendix
Table A1—Income Groups by 1990 GDP per Capita
| Group | Countries | Members |
|---|---|---|
| High | 40 | Australia, Austria, Belgium, Canada, Cyprus, Czechia†, Denmark, Estonia†, Finland, France, Germany, Greece, Hong Kong, Ireland, Israel, Italy, Japan, Kuwait*, Latvia†, Lithuania†, Luxembourg, Malta, Netherlands, New Zealand, Norway, Portugal, Saudi Arabia*, Singapore, Slovakia†, Slovenia†, South Korea, Spain, Sweden, Switzerland, Taiwan, Trinidad and Tobago*, United Arab Emirates*, United Kingdom, United States, Venezuela* |
| Middle | 39 | Argentina, Armenia, Azerbaijan, Bahrain, Belarus, Botswana, Brazil, Bulgaria, Chile, Colombia, Costa Rica, Croatia, Ecuador, Georgia, Guatemala, Hungary, Iran, Jamaica, Jordan, Kazakhstan, Kyrgyzstan, Malaysia, Mauritius, Mexico, Namibia, Oman, Panama, Paraguay, Peru, Poland, Qatar, Romania, Russia, South Africa, Thailand, Tunisia, Turkey, Ukraine, Uruguay |
| Low | 40 | Albania, Bangladesh, Benin, Bolivia, Burkina Faso, Cambodia, Cameroon, China, Cote d'Ivoire, Dominican Republic, Egypt, El Salvador, Ethiopia, Ghana, Guinea, Honduras, India, Indonesia, Kenya, Laos, Madagascar, Malawi, Mongolia, Morocco, Mozambique, Nepal, Nicaragua, Nigeria, Pakistan, Philippines, Rwanda, Senegal, Sri Lanka, Tajikistan, Tanzania, Togo, Uganda, Vietnam, Zambia, Zimbabwe |
| High: net exporters of embodied CO₂, 1990 | 10 | Australia, Belgium, Czechia, Estonia, Luxembourg, Saudi Arabia, Taiwan, Trinidad and Tobago, United States, Venezuela (Kuwait and Norway have no 1990 value in the sample) |
| High: net exporters of embodied CO₂, 1997 | 12 | Australia, Belgium, Cyprus, Czechia, Estonia, Kuwait, Luxembourg, Saudi Arabia, Slovakia, Trinidad and Tobago, United States, Venezuela (Norway has no 1997 value) |
| High: net exporters of embodied CO₂, 2007 | 7 | Australia, Czechia, Kuwait, Luxembourg, Saudi Arabia, Trinidad and Tobago, Venezuela |
Notes: Country names as in the file. * Oil exporter, by an outside criterion (exports dominated by oil and gas) that Norway also meets; † transition economy. Cut-offs are 4,695 and 12,349 international dollars at 2011 prices.
Table A2—Annual First-Difference Elasticities with Group-by-Period Intercepts (Descriptive)
| (1) | (2) | (3) | |
|---|---|---|---|
| × Low × pre | 0.532*** (0.102) [0.207] | 0.383*** (0.088) [0.203] | −0.149 (0.106) [0.078] |
| × Low × post | 0.987*** (0.156) [0.137] | 1.086*** (0.231) [0.146] | 0.099 (0.132) [0.142] |
| × Middle × pre | 0.627*** (0.099) [0.082] | 0.573*** (0.081) [0.096] | −0.055 (0.124) [0.071] |
| × Middle × post | 0.859*** (0.143) [0.142] | 0.683*** (0.203) [0.121] | −0.176 (0.269) [0.169] |
| × High × pre | 0.661*** (0.179) [0.132] | 0.654*** (0.103) [0.090] | −0.007 (0.208) [0.134] |
| × High × post | 0.577*** (0.066) [0.068] | 0.567 (0.427) [0.246] | −0.010 (0.430) [0.219] |
| Observations / countries | 3,782 / 119 | 3,782 / 119 | 3,782 / 119 |
| R² | 0.201 | 0.083 | 0.016 |
| High wedge without group-by-period intercepts (pre / post) | −0.014 (0.136) / −0.014 (0.419) |
Notes: Equation (4), regressor , with year fixed effects and group-by-period intercepts; consecutive years only; pre is 1991–2007 and post 2008–2022. Driscoll–Kraay standard errors (two lags) are in parentheses, with stars based on them; country-clustered standard errors are in brackets. With only about 15 post-2007 years, Driscoll–Kraay inference is itself fragile. Outside benchmark years the annual consumption series is built from GDP and trade time series, so annual slopes largely reflect construction; the table is descriptive, not robustness evidence. Kuwait 1990–1992 is excluded; without group-by-period intercepts and with those years kept, the high × pre wedge coefficient is 1.475 (0.428) [0.551]. * p<0.10, ** p<0.05, *** p<0.01.
Table A3—File-Internal Test of the Benchmark Years
| Statistic | Benchmark-year mean | Other-year mean | Permutation p-value | Benchmark-year ranks (of 30; 1 = largest) |
|---|---|---|---|---|
| Kink in the annual wedge change (median absolute second difference across countries) | 0.051 | 0.060 | 0.831 | 1997: 21; 2001: 14; 2004: 23; 2007: 15; 2011: 8; 2014: 16; 2017: 24 |
| Cross-country standard deviation of the annual wedge change | 0.141 | 0.144 | 0.540 | 1997: 18; 2001: 9; 2004: 25; 2007: 13; 2011: 3; 2014: 16; 2017: 27 |
Notes: The test asks whether the benchmark years named in the Global Carbon Budget 2025 supplement (1997, 2001, 2004, 2007, 2011, 2014, 2017) appear in the file as kinks or jumps in dispersion of the wedge path, over 1992–2021. The permutation test draws 7 of the 30 years at random (20,000 draws); the p-value is the share of draws whose mean is at least the benchmark-year mean. The five years with the largest kinks are 1999, 2000, 2008, 2009 and 1998; the five with the largest dispersion are 2000, 2018, 2011, 2003 and 1996. The test's power is unknown, and its largest kinks coincide with the shocks of 1998–2000 and 2008–2009, so its null neither confirms nor refutes the benchmark dating.
Table A4—Shares of the Top Tercile's Territorial CO₂ (Percent)
| 1990 | 1997 | 2007 | 2017 | 2022 | |
|---|---|---|---|---|---|
| Members observed | 38 | 39 | 40 | 40 | 40 |
| Oil exporters and transition economies | 6.0 | 5.8 | 7.6 | 10.4 | 11.0 |
| Oil exporters | 3.2 | 4.0 | 6.0 | 8.8 | 9.6 |
| Transition economies | 2.8 | 1.8 | 1.6 | 1.6 | 1.4 |
| Oil exporters, transition economies and Norway | 6.0 | 5.8 | 7.9 | 10.8 | 11.4 |
| United States | 43.9 | 44.6 | 43.7 | 41.0 | 42.7 |
| Members outside the 1997 net importers | 52.4 | 53.2 | 54.3 | 54.1 | 56.2 |
| Members outside the 1990 net importers | 52.9 | 55.9 |
Notes: The denominator is, in each year, the summed co2 of the members of the top tercile (defined by 1990 GDP per capita) observed in the analysis sample that year, so Kuwait 1990–1992 and Norway before 2003 are excluded. Norway's share is 0.3 percent in 2007, 2017 and 2022.
Table A5—Rules from the File's Columns for the Oil-Exporter Set
| Rule | Threshold keeping all five | Countries selected | Norway selected (value; rank) | Others selected beyond the five |
|---|---|---|---|---|
Flaring share of co2, mean 1990–2022 | 1.36% | 9 | Yes (3.94%; 2 of 40) | Australia, Canada, New Zealand |
| Flaring per person, mean 1990–2022 | 0.263 t | 8 | Yes (0.354 t; 6 of 40) | Australia, Canada |
Flaring share of co2, 1990 | 0.25% | 20 | Yes (5.47%; 3 of 37) | Australia, Canada and 12 others |
Flaring share of co2, 1997 | 1.63% | 9 | Yes (4.66%; 2 of 40) | Australia, Canada, New Zealand |
Flaring share of co2, 2007 | 0.76% | 15 | Yes (6.18%; 1 of 40) | Australia, Canada and 7 others |
| Net exporter of embodied CO₂, 1997 | trade_co2 ≤ 0 | 12 | No (no value) | Misses the United Arab Emirates; adds 8 |
| Net exporter of embodied CO₂, 2007 | trade_co2 ≤ 0 | 7 | No | Misses the United Arab Emirates; adds Australia, Czechia, Luxembourg |
Notes: The five are Kuwait, Venezuela, Trinidad and Tobago, the United Arab Emirates and Saudi Arabia. Flaring rules use the lowest threshold that keeps all five, over the 40 top-tercile countries and all file rows for 1990–2022 (37 countries have a 1990 flaring value). No rule reproduces the set; the file has no oil production or export column.
Table A6—Slopes That Pass the Inference Rule: Leaving Out One Country at a Time
| Slope | N | Full-sample slope (wild p) | Leave-one-out range (country dropped) | Most influential country: slope without it (HC3 p; wild p) | Largest leverage (value) | Largest absolute studentized residual (value) | Largest share of robust variance term (%) | Leave-one-out wild p range; fits with p < 0.10 |
|---|---|---|---|---|---|---|---|---|
| Core, 2007–2017 | 29 | 0.985 (0.068) | [0.304 (Malta), 1.265 (Greece)] | Malta: 0.304 (0.091; 0.068) | Greece (0.312) | Malta (4.43) | Malta (76.1) | [0.047, 0.148]; 27 of 29 |
| Core without Norway, 2007–2017 | 28 | 1.024 (0.051) | [0.339 (Malta), 1.316 (Singapore)] | Malta: 0.339 (0.066; 0.044) | Greece (0.312) | Malta (4.37) | Malta (74.4) | [0.035, 0.109]; 26 of 28 |
| Net importers (1997 sign), 1997–2007 | 27 | 0.279 (0.091) | [0.203 (Singapore), 0.512 (United Arab Emirates)] | United Arab Emirates: 0.512 (0.059; 0.069) | United Arab Emirates (0.585) | Singapore (2.83) | United Arab Emirates (44.2) | [0.069, 0.144]; 15 of 27 |
| Net importers (1990 sign), 1997–2007 | 28 | 0.294 (0.097) | [0.223 (Singapore), 0.562 (United Arab Emirates)] | United Arab Emirates: 0.562 (0.028; 0.036) | United Arab Emirates (0.574) | Singapore (3.00) | United Arab Emirates (54.3) | [0.036, 0.160]; 5 of 28 |
Notes: Each fit drops one top-group country and re-estimates the wedge slope of equation (2). The most influential country is the one whose removal changes the slope most. Leverage, studentized residuals and Cook's distance come from OLS on the group's subsample; mean leverage is , about 0.07. The robust variance term is , the numerator of the slope's heteroskedasticity-robust variance. Wild-bootstrap seeds are those of the corresponding Table 4 rows. In the core group, Malta's 2007–2017 change in log GDP is 0.454 and in the wedge 1.501, with leverage 0.175 and Cook's distance 2.08; without Malta the core slope's HC1 p-value is 0.069. In the net-importer groups the United Arab Emirates' 1997–2007 change in log GDP is 1.254, and its Cook's distance is 0.97 and 1.58. Malta is in both net-importer groups; without it their slopes are 0.298 and 0.313, with wild-bootstrap p-values of 0.0996 and 0.1035. Diagnostic, added after the main results were known.
Table A7—Multiple Testing of the Top Group's Wedge Slopes
| Family of tests | Tests | Wild p < 0.10 | Expected if every null holds | Probability of at least as many if independent | Smallest wild p | Bonferroni threshold (10%) | Significant after Bonferroni / Holm |
|---|---|---|---|---|---|---|---|
| Table 4, unweighted | 20 | 4 | 2.0 | 0.133 | 0.051 | 0.005 | 0 / 0 |
| Table 4, weighted | 20 | 0 | 2.0 | 1.000 | 0.140 | 0.005 | 0 / 0 |
| All long-difference wedge slopes in Tables 3–6, unweighted | 29 | 4 | 2.9 | 0.329 | 0.051 | 0.0034 | 0 / 0 |
Notes: The core group without Norway's 1990–2007 and 1997–2007 rows equal the core rows and are counted once; the two Kuwait treatments in Table 5 give the same 1990–2007 estimate and count as one test; the two benchmark-to-benchmark rows of Table 6 are the full-group rows of Table 4. The tests are nested in time and overlap in sample, so they are not independent, and the independent-case probability is only a reference. Holm's step-down correction also rejects nothing at 5 percent. Diagnostic, added after the main results were known.
Table A8—Top-Group Long-Difference Wedge Slopes without Malta
| Sample | Period | N | Full sample: slope (HC3 p; wild p) | Without Malta: slope (HC3 p; wild p) | Malta's share of robust variance term (%) | Malta's studentized residual | Malta's Cook's distance |
|---|---|---|---|---|---|---|---|
| All top tercile | 1990–2007 | 38 | 0.0274 (0.8421; 0.8086) | 0.0259 (0.8542; 0.8335) | 0.0 | 0.88 | 0.01 |
| 2007–2022† | 40 | 0.1866 (0.5048; 0.5005) | −0.0075 (0.8961; 0.8847) | 73.8 | 4.78 | 1.02 | |
| 1997–2007 | 39 | 0.0574 (0.6479; 0.6125) | 0.0740 (0.5535; 0.5159) | 1.9 | 1.26 | 0.03 | |
| 2007–2017† | 40 | 0.4370 (0.2486; 0.2127) | 0.1247 (0.3728; 0.3480) | 72.2 | 5.15 | 1.27 | |
| 2017–2022 | 40 | −0.0962 (0.7322; 0.5581) | −0.0976 (0.7503; 0.5427) | 0.0 | 0.07 | 0.00 | |
| Core | 1990–2007 | 28 | 0.3055 (0.2998; 0.3367) | 0.2974 (0.3256; 0.3639) | 0.1 | 0.87 | 0.02 |
| 2007–2022† | 29 | 0.7383 (0.2758; 0.2132) | 0.1018 (0.5305; 0.5048) | 86.7 | 3.90 | 2.11 | |
| 1997–2007 | 28 | 0.3250 (0.3898; 0.3646) | 0.3481 (0.3590; 0.3307) | 0.4 | 1.20 | 0.03 | |
| 2007–2017† | 29 | 0.9852 (0.2181; 0.0675) | 0.3039 (0.0912; 0.0684) | 76.1 | 4.43 | 2.08 | |
| 2017–2022 | 29 | −0.5608 (0.2549; 0.2547) | −0.6507 (0.2474; 0.2548) | 3.1 | 0.59 | 0.04 | |
| Core, also without Norway | 2007–2022† | 28 | 0.7809 (0.2437; 0.1213) | 0.1576 (0.2451; 0.2326) | 84.9 | 3.98 | 2.25 |
| 2007–2017† | 28 | 1.0243 (0.2092; 0.0514) | 0.3387 (0.0656; 0.0436) | 74.4 | 4.37 | 2.10 | |
| 2017–2022 | 28 | −0.6090 (0.1988; 0.1571) | −0.6883 (0.2049; 0.1956) | 2.6 | 0.67 | 0.05 | |
| Net importers, 1997 sign | 1990–2007 | 27 | 0.1303 (0.3440; 0.1831) | 0.1282 (0.3461; 0.2067) | 0.0 | 0.78 | 0.01 |
| 2007–2022† | 27 | 0.5270 (0.3783; 0.4849) | 0.0208 (0.9189; 0.9176) | 73.6 | 3.97 | 1.70 | |
| 1997–2007 | 27 | 0.2786 (0.2845; 0.0908) | 0.2983 (0.2682; 0.0996) | 1.7 | 1.42 | 0.05 | |
| 2007–2017† | 27 | 0.6256 (0.3406; 0.3462) | 0.1435 (0.5351; 0.4846) | 64.0 | 4.34 | 1.42 | |
| 2017–2022 | 27 | −0.4263 (0.3846; 0.5308) | −0.4585 (0.4209; 0.5892) | 0.4 | 0.27 | 0.01 | |
| Net importers, 1990 sign | 1990–2007 | 28 | 0.1049 (0.4204; 0.3221) | 0.1024 (0.4393; 0.3475) | 0.0 | 0.90 | 0.02 |
| 1997–2007 | 28 | 0.2935 (0.3125; 0.0971) | 0.3127 (0.2982; 0.1035) | 1.4 | 1.47 | 0.05 | |
| Table 5: regressor | 2007–2022† | 40 | 0.2080 (0.5264; 0.4211) | 0.0254 (0.6558; 0.5419) | 70.6 | 4.78 | 0.79 |
| Table 6: averaged endpoints | 2005–07 → 2020–22† | 40 | 0.0962 (0.7096; 0.8760) | −0.0906 (0.1414; 0.1758) | 73.2 | 5.25 | 0.97 |
| Table 6: 2007 → 2019 | 2007–2019† | 40 | 0.5355 (0.0810; 0.2719) | 0.2856 (0.4019; 0.5920) | 72.4 | 5.05 | 1.24 |
Notes: Unweighted slopes of equation (2), as in Tables 4–6; the core group without Norway's 1990–2007 and 1997–2007 rows equal the core rows. Each Malta-deleted fit uses the wild-bootstrap seed of its full-sample row (Webb weights, null imposed, 9,999 draws). Malta's territorial CO₂ is 2.728 Mt in 2007 and 1.774 Mt in 2022, and its consumption-based CO₂ 5.128 and 16.119 Mt, so its wedge rose 1.575 on log GDP growth of 0.669. Diagnostics are Malta's in the full sample; the robust variance term is . † Malta has the largest share of the robust variance term. In the other long-difference rows of Tables 5 and 6 Malta's share is below 1 percent and the slope moves by less than 0.01; the population ≥ 1 million samples exclude Malta. In the weighted regressions of Tables 3, 4 and 6 Malta's share is at most 2.4 percent, no weighted slope moves by more than 0.016 (Table 3: 0.0663 and −0.0465 without Malta), and none passes the rule with or without Malta; Table 3's weighted-minus-unweighted differences become 0.040 (0.129) and −0.039 (0.063). Removing Malta changes the aggregate wedge changes of Table 4 by at most 0.002. As a sensitivity check, a screen that drops, in every sample alike, each country with Cook's distance above 4/n changes which slopes pass rather than settling anything: 8 of the 52 distinct top-group long-difference slopes, unweighted and weighted, then pass, four before and four after 2007, and only one of the four that pass in the full samples (the 1990-sign net importers, 1997–2007). In the full group's and the 1997 net importers' 2007–2017 samples the screen drops exactly Malta and the United Arab Emirates, giving slopes of 0.228 (HC3 p = 0.033; wild-bootstrap p = 0.037) and 0.344 (HC3 p = 0.049; wild-bootstrap p = 0.032), so the small Malta-deleted slopes of 0.125 and 0.144 depend on keeping the Emirates. We draw no conclusion from it. Diagnostic, added after the main results were known.
7.Review
Automated review: Accept (novelty 2, rigor 4, clarity 4, reproducibility 4).
| Novelty 2/5 | Rigor 4/5 | Clarity 4/5 | Reproducibility 4/5 |
Refereed by Claude Opus 5.5, the model that also wrote the paper.
This confirmation round checks the two text changes from my last report against revision.round6 and round7 of the record. Both are made correctly. First, the Introduction and the Conclusion now report that the net-importer passes are fragile: the UAE deletion strengthens them, they pass in only 15 of 27 and 5 of 28 single-country deletions, and the 1990-sign pass fails without Malta (wild p 0.104). Second, the Emirates sensitivity is limited to the Malta-deleted 2007–2017 slopes of the full group and the 1997 net importers, where the Cook's-distance screen drops exactly Malta and the UAE. Every quoted number matches the record, nothing else changed or regressed, and the abstract is exactly 100 words. The remaining limitation, modest novelty for a descriptive study without exogenous variation, was set aside in earlier rounds and is disclosed in the paper. The claims now match the evidence. I recommend acceptance.
Strengths
- Both changes I asked for were made, and both match the record. In the Introduction and the Conclusion, 'All four still pass when their most influential country is deleted' is now followed by the fragility that Section IV reports. Removing the United Arab Emirates raises both net-importer slopes and lowers their p-values, so surviving that deletion says little about robustness. The slopes pass in only 15 of 27 and 5 of 28 single-country deletions, and the 1990-sign slope fails without Malta (wild-bootstrap p = 0.104). These numbers match revision.round6.leave_one_out_passing: n_loo_fits_sig10_wild is 15 of 27 for the 1997 sign and 5 of 28 for the 1990 sign, and loo.Malta p_wild is 0.1035.
- The Emirates sensitivity is now limited to the 2007–2017 slopes in both the Introduction and the Conclusion, as it already was in Section IV and the Table A8 note. This matches revision.round7.table4. In the 2007–2017 unweighted samples of the full group and the 1997 net importers, the Cook's-distance screen drops exactly Malta and the United Arab Emirates, giving 0.2275 (wild p 0.0371) and 0.3441 (wild p 0.0320). In the full group's 2007–2022 sample the screen drops Malta and Venezuela instead, and the paper no longer implies otherwise.
- Nothing regressed. The script and RESULT_JSON are unchanged. The leave-one-out slopes quoted alongside the revised sentences match round6: without Malta the core slopes are 0.3039 and 0.3387 (HC3 p 0.0912 and 0.0656; wild p 0.0684 and 0.0436), and without the UAE the net-importer slopes are 0.5116 and 0.5617 (HC3 p 0.0590 and 0.0275; wild p 0.0687 and 0.0360). The parts of the paper I accepted last round are unchanged: the period-comparison wording, the post hoc labels, the provenance statement (OWID named only as the user's statement), the Kuwait disclosure and the multiple-testing results. By whitespace count the abstract is still exactly 100 words, within the limit.
Weaknesses
- This is unchanged from earlier rounds and is not counted against this revision. Novelty is modest for AER: the study is descriptive, has no exogenous variation, and neither pre-specified growth slope is informative. The paper says all of this plainly.
Required changes
—
Record
§Event timeline
- Run createdQuestion: “'No question given: what can be learned from the user-supplied data file owid-co2-data.csv (50,411 rows; columns: country, year, iso_code, population, gdp, cement_co2, cement_co2_per_capita, co2, …)' language=en discipline=economics venue='American Economic Review' mode=user_data” · output language: English · models: agent:claude-opus-5-5
- Data suppliedowid-co2-data.csv · 50,411 rows · 79 columns · 32 MB · sha256 7ec10de1e950… · no question given
- Stage started · Data profileAttempt 1 · agent:claude-opus-5-5
- Data profiled · Data profile50,411 rows × 79 columns · every figure exact
- awaiting_agent · Data profilecomplete_json response for data_profile: data_profile-complete_json-1.response.json
- Resume requested
- Stage started · Data profileAttempt 2 · agent:claude-opus-5-5
- Data profiled · Data profile50,411 rows × 79 columns · every figure exact
- awaiting_agent · Data profilecomplete_json response for data_profile: data_profile-complete_json-1.response.json
- Resume requested
- Stage started · Data profileAttempt 3 · agent:claude-opus-5-5
- Data profiled · Data profile50,411 rows × 79 columns · every figure exact
- Awaiting approval · Data profile
- Sent back for revision · Data profileRe-running from Data profile: “Mostly right; please correct these points. (1) iso_code: the 36 entities without a code include Kosovo and Ryukyu Islands, so say 34 aggregates/non-territorial entities plus those two places, not '36 ”
- Stage started · Data profileAttempt 4 · agent:claude-opus-5-5
- Data profiled · Data profile50,411 rows × 79 columns · every figure exact
- awaiting_agent · Data profilecomplete_json response for data_profile: data_profile-complete_json-1.response.json
- Resume requested
- Stage started · Data profileAttempt 5 · agent:claude-opus-5-5
- Data profiled · Data profile50,411 rows × 79 columns · every figure exact
- awaiting_agent · Data profilecomplete_json response for data_profile: data_profile-complete_json-1.response.json
- Resume requested
- Stage started · Data profileAttempt 6 · agent:claude-opus-5-5
- Data profiled · Data profile50,411 rows × 79 columns · every figure exact
- Awaiting approval · Data profile
- Approved · Data profile
- Description confirmed · Data profileConfirmed by the owner
- Stage started · Literature reviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Literature reviewsearch response for literature_review: literature_review-search-1.response.json
- Resume requested
- Stage started · Literature reviewAttempt 2 · agent:claude-opus-5-5
- awaiting_agent · Literature reviewsearch response for literature_review: literature_review-search-1.response.json
- Resume requested
- Stage started · Literature reviewAttempt 3 · agent:claude-opus-5-5
- Search complete · Literature review43 sources · 11,369 characters of notes
- awaiting_agent · Literature reviewcomplete_json response for literature_review: literature_review-complete_json-1.response.json
- Resume requested
- Stage started · Literature reviewAttempt 4 · agent:claude-opus-5-5
- Search complete · Literature review43 sources · 11,369 characters of notes
- Stage completed · Literature review
- Stage started · IdeationAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Ideationcomplete_json response for ideation: ideation-complete_json-1.response.json
- Resume requested
- Stage started · IdeationAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Ideation
- Approved · IdeationIdea 1 selected
- Question rewordedPreviously: “No question given: what can be learned from the user-supplied data file owid-co2-data.csv (50,411 rows; columns: country, year, iso_code, population, gdp, cement_co2, cement_co2_per_capita, co2, …)” · no question was given; the approved idea defines it
- Stage started · ExperimentationAttempt 1 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 2 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 3 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- Experiment round · ExperimentationRound 1: exit 0; result parsed (1.01s)
- Stage completed · Experimentation
- Stage started · AnalysisAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 2 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 2 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Sent back for revision · ReviewRe-running from Experimentation: “Independent referee (Claude Opus 5.5): major_revision. Required changes: (1) [text] In Section III, delete 'Before estimating anything'. State that an exploratory run made before the plan was fixed ke”
- Stage started · ExperimentationAttempt 4 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 5 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- Experiment round · ExperimentationRound 1: exit 0; result parsed (3.827s)
- Stage completed · Experimentation
- Stage started · AnalysisAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 2 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 3 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 4 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Sent back for revision · ReviewRe-running from Experimentation: “Independent referee (Claude Opus 5.5), round 2: minor_revision. Required changes: (1) [analysis] Add to Table 2 a 2007-2017 benchmark-to-benchmark row (and, if convenient, 2007-2019) for the top terci”
- Stage started · ExperimentationAttempt 6 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 7 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- Experiment round · ExperimentationRound 1: exit 0; result parsed (3.879s)
- Stage completed · Experimentation
- Stage started · AnalysisAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 2 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 5 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 6 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Approved · Review
- Run completereport=/Users/haodingresearch/Desktop/auto-research/runs/studies-data/en/runs/f0103331d627/REPORT.md
- Reopened after sign-off · ReviewReopened after sign-off; re-running from Experimentation: “Editor's consistency revision. The Chinese edition of this study (same file, same script), at its independent referee's request, extended the shared script: the top group's membership, the group witho”
- Stage started · ExperimentationAttempt 8 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 9 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 10 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- Experiment round · ExperimentationRound 1: exit 0; result parsed (5.211s)
- Stage completed · Experimentation
- Stage started · AnalysisAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 2 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 7 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 8 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Sent back for revision · ReviewRe-running from Experimentation: “Independent referee (Claude Opus 5.5), reopened round: minor_revision. Required changes: (1) [analysis] Compute in the script, and report in RESULT_JSON, the oil-exporter-plus-transition and United St”
- Stage started · ExperimentationAttempt 11 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 12 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- Experiment round · ExperimentationRound 1: exit 0; result parsed (5.318s)
- Stage completed · Experimentation
- Stage started · AnalysisAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 2 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 9 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 10 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Sent back for revision · ReviewRe-running from Analysis: “Independent referee (Claude Opus 5.5): minor_revision. Required changes: (1) [text] In the Introduction and Conclusion, list all four top-group slopes that pass the wild-bootstrap rule: core 2007-2017”
- Stage started · AnalysisAttempt 3 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 4 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 5 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 11 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Approved · Review
- Run completereport=/Users/haodingresearch/Desktop/auto-research/runs/studies-data/en/runs/f0103331d627/REPORT.md
- Reopened after sign-off · ReviewReopened after sign-off; re-running from Experimentation: “Editor's consistency revision. The shared analysis script has been extended to v6 (new key revision.round6; every earlier result unchanged): leave-one-out re-estimates of the four slopes that pass the”
- Stage started · ExperimentationAttempt 13 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 14 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- Experiment round · ExperimentationRound 1: exit 0; result parsed (5.765s)
- Stage completed · Experimentation
- Stage started · AnalysisAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 2 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 3 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 12 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 13 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Sent back for revision · ReviewRe-running from Experimentation: “Independent referee (Claude Opus 5.5): minor_revision. Required changes: (1) [analysis] Extend the Malta deletion to every top-group post-2007 wedge slope in Tables 4-6 and report each slope, its HC3 ”
- Stage started · ExperimentationAttempt 15 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 16 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- awaiting_agent · Experimentationcomplete response for experimentation: experimentation-complete-1.response.txt
- Resume requested
- Stage started · ExperimentationAttempt 17 · agent:claude-opus-5-5
- Dataset provisioned · Experimentationowid-co2-data.csv (your data) · 50,411 observations · 79 variables · sha256 7ec10de1e950
- Experiment round · ExperimentationRound 1: exit 0; result parsed (6.978s)
- Stage completed · Experimentation
- Stage started · AnalysisAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 2 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 14 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 15 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Sent back for revision · ReviewRe-running from Analysis: “Independent referee (Claude Opus 5.5): accept, with two wording points the editor adopts. (1) Malta dominates every unweighted sample that starts in 2007 and includes Malta; the sample restricted to p”
- Stage started · AnalysisAttempt 3 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 4 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 5 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 16 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 3 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 4 · agent:claude-opus-5-5
- Awaiting approval · Review
- Sent back for revision · ReviewRe-running from Paper writing: “Editor: the abstract is 101 words, one over the venue's 100-word limit. Trim 'Without one country, Malta, the' to 'Without Malta, the'. Change nothing else.”
- Stage started · Paper writingAttempt 17 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 18 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Approved · Review
- Run completereport=/Users/haodingresearch/Desktop/auto-research/runs/studies-data/en/runs/f0103331d627/REPORT.md
- Reopened after sign-off · ReviewReopened after sign-off; re-running from Analysis: “Editor's consistency revision (text only; script and numbers unchanged). (1) Restrict 'the ordering reverses in every group' to the 1990-2007 vs 2007-2022 comparison; on the benchmark endpoints withou”
- Stage started · AnalysisAttempt 6 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 7 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 8 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 19 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 20 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Sent back for revision · ReviewRe-running from Analysis: “Independent referee (Claude Opus 5.5): minor_revision. Required changes: (1) [text] In the Introduction and the Conclusion, follow 'All four still pass when their most influential country is deleted' ”
- Stage started · AnalysisAttempt 9 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 10 · agent:claude-opus-5-5
- awaiting_agent · Analysiscomplete_json response for analysis: analysis-complete_json-1.response.json
- Resume requested
- Stage started · AnalysisAttempt 11 · agent:claude-opus-5-5
- Stage completed · Analysis
- Stage started · Paper writingAttempt 21 · agent:claude-opus-5-5
- awaiting_agent · Paper writingcomplete response for paper_writing: paper_writing-complete-1.response.txt
- Resume requested
- Stage started · Paper writingAttempt 22 · agent:claude-opus-5-5
- Stage completed · Paper writing
- Stage started · ReviewAttempt 1 · agent:claude-opus-5-5
- awaiting_agent · Reviewcomplete_json response for review: review-complete_json-1.response.json
- Resume requested
- Stage started · ReviewAttempt 2 · agent:claude-opus-5-5
- Awaiting approval · Review
- Approved · Review
- Run completereport=/Users/haodingresearch/Desktop/auto-research/runs/studies-data/en/runs/f0103331d627/REPORT.md